Files
nexus/sreweekly/articles/400/07-how-does-generative-ai-work-with-devops-and-incident-response.html
2026-09-12 17:23:01 +08:00

187 lines
365 KiB
HTML
Raw Blame History

This file contains ambiguous Unicode characters
This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.
<!DOCTYPE html><html lang="en"><head><script>
window.VIDEOJS_NO_DYNAMIC_STYLE = true;
window.ENV = {"SANITY_DATASET":"production","ALGOLIA_INDEX":"heavybit-library","ALGOLIA_APPLICATION_ID":"0QCKBC1FMK","ALGOLIA_SITEWIDE_INDEX":"heavybit-sitewide","ALGOLIA_JOBS_INDEX":"heavybit-jobs","MUX_ENV_KEY":"dRI98bL8ze03gYDL6-Ogle2Js","ALGOLIA_SEARCH_API_KEY":"7733fdd199ec7af7ee2923ec39c3f036","MODE":"production","VITE_CONTEXT":"production"};
</script><meta theme-color="#556F73"/><meta apple-mobile-web-app-status-bar-style="#556F73"/><meta charSet="utf-8"/><meta name="viewport" content="width=device-width,initial-scale=1"/><link rel="stylesheet" href="/assets/root-DDCa6fA8.css"/><title>How Does Generative AI Work with Devops and Incident Response? | Heavybit</title><meta property="og:title" name="og:title" content="How Does Generative AI Work with Devops and Incident Response? | Heavybit"/><meta property="twitter:title" name="twitter:title" content="How Does Generative AI Work with Devops and Incident Response? | Heavybit"/><link rel="canonical" href="https://www.heavybit.com/library/article/generative-ai-incident-response-devops"/><meta property="og:url" name="og:url" content="https://www.heavybit.com/library/article/generative-ai-incident-response-devops"/><meta property="description" name="description" content="DevOps and IM experts from Jeli, PagerDuty, AWS, and other leading outfits explain how generative AI will affect the future of site reliability engineering."/><meta property="og:description" content="DevOps and IM experts from Jeli, PagerDuty, AWS, and other leading outfits explain how generative AI will affect the future of site reliability engineering."/><meta property="og:image" name="og:image" content="https://cdn.sanity.io/images/50q6fr1p/production/b67963a5a553d32ba4c307eb7425f753bcdaba6a-800x300.jpg?auto=format&amp;dpr=2"/><meta property="twitter:card" name="twitter:card" content="summary_large_image"/><meta content="Heavybit" property="og:site_name"/><meta content="@heavybit" name="twitter:site"/><meta content="@heavybit" name="twitter:creator"/><link rel="sitemap" type="application/xml" title="Sitemap" href="https://www.heavybit.com/sitemap.xml"/><link rel="icon" href="/icons/favicon-dark.ico"/></head><body class="min-h-screen lining-nums undefined bg-darker-steel-gray subpixel-antialiased"><noscript><iframe src="https://www.googletagmanager.com/ns.html?id=GTM-MH5753" height="0" width="0" style="display:none;visibility:hidden"></iframe></noscript><div class="bg-heavy-slate safari-full fixed z-[-5] h-screen w-full"></div><header class="bg-heavy-slate border-darker-steel-gray relative z-50 border-b border-solid"><div class="mx-auto max-w-(--breakpoint-xl) px-6"><div class="relative grid h-[90px] box-border w-full grid-cols-3 justify-items-center items-center py-[30px]"><div class="relative max-h-full h-full flex items-center place-self-start"><button class="block text-2xl text-white xl:hidden"><svg data-prefix="fas" data-icon="bars" class="svg-inline--fa fa-bars " role="img" viewBox="0 0 448 512" aria-hidden="true"><path fill="currentColor" d="M0 96C0 78.3 14.3 64 32 64l384 0c17.7 0 32 14.3 32 32s-14.3 32-32 32L32 128C14.3 128 0 113.7 0 96zM0 256c0-17.7 14.3-32 32-32l384 0c17.7 0 32 14.3 32 32s-14.3 32-32 32L32 288c-17.7 0-32-14.3-32-32zM448 416c0 17.7-14.3 32-32 32L32 448c-17.7 0-32-14.3-32-32s14.3-32 32-32l384 0c17.7 0 32 14.3 32 32z"></path></svg></button><ul class="absolute top-0 -mt-[5px] -ml-[20px] hidden w-full gap-x-[0px] xl:flex"><li class="group relative rounded-[3px]"><a class="block rounded-t-[3px] bg-transparent group-hover:bg-white" title="Navigate to about page" href="/about" data-discover="true"><span class="group-hover:text-darker-steel-gray flex items-center rounded-t-[3px] bg-transparent px-[20px] py-[15px] uppercase text-dark-steel-gray hover:bg-light-steel-gray font-plexmono text-caption2">About<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a><ul class="relative z-100 h-0 overflow-hidden rounded-b-[3px] bg-white group-hover:block group-hover:h-min"><li><a href="/team" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Team<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/faq" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">FAQ<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/press" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">News<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li></ul></li><li class="group relative rounded-[3px]"><a class="block rounded-t-[3px] bg-transparent group-hover:bg-white" title="Navigate to portfolio page" href="/portfolio" data-discover="true"><span class="group-hover:text-darker-steel-gray flex items-center rounded-t-[3px] bg-transparent px-[20px] py-[15px] uppercase text-dark-steel-gray hover:bg-light-steel-gray font-plexmono text-caption2">Portfolio<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a><ul class="relative z-100 h-0 overflow-hidden rounded-b-[3px] bg-white group-hover:block group-hover:h-min"><li><a href="/portfolio/spotlights" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Spotlights<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/jobs" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Jobs<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li></ul></li><li class="group relative rounded-[3px]"><a class="block rounded-t-[3px] bg-transparent group-hover:bg-white" title="Navigate to community page" href="/community" data-discover="true"><span class="group-hover:text-darker-steel-gray flex items-center rounded-t-[3px] bg-transparent px-[20px] py-[15px] uppercase text-dark-steel-gray hover:bg-light-steel-gray font-plexmono text-caption2">Community<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a><ul class="relative z-100 h-0 overflow-hidden rounded-b-[3px] bg-white group-hover:block group-hover:h-min"><li><a href="/events" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Events<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/community/advisors" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Advisors<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/devguild" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">DevGuild<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li></ul></li><li class="group relative rounded-[3px]"><a class="block rounded-t-[3px] bg-transparent group-hover:bg-white" title="Navigate to library page" href="/library" data-discover="true"><span class="group-hover:text-darker-steel-gray flex items-center rounded-t-[3px] bg-transparent px-[20px] py-[15px] uppercase text-dark-steel-gray hover:bg-light-steel-gray font-plexmono text-caption2">Library<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a><ul class="relative z-100 h-0 overflow-hidden rounded-b-[3px] bg-white group-hover:block group-hover:h-min"><li><a href="/library?filter=articles" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Articles<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/library?filter=videos" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Videos<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/library/podcasts" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Podcast Network<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/library/collections" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Collections<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/reports" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">Reports<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li><li><a href="/devtoolsdigest" class="hover:bg-light-steel-gray block px-[20px] py-[15px]" data-discover="true"><span class="text-darker-steel-gray text-caption2 font-plexmono flex items-center whitespace-nowrap uppercase hover:opacity-100">DevToolsDigest<svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right opacity-0 text-orange ml-[8px] transition-all" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></span></a></li></ul></li></ul></div><div><a class="mx-5 block" href="/" data-discover="true"><svg width="43" height="37" viewBox="0 0 43 37" fill="none" xmlns="http://www.w3.org/2000/svg"><path fill-rule="evenodd" clip-rule="evenodd" d="M0.359985 18.4985L10.9301 37H32.0734L42.6435 18.4985L32.0734 0H10.9301L0.359985 18.4985ZM34.7152 18.4985L28.1108 30.0623H14.8958L8.28834 18.4985L14.8958 6.93769H28.1108L34.7152 18.4985Z" fill="white"></path><path fill-rule="evenodd" clip-rule="evenodd" d="M31.4128 17.3427L26.7775 9.24924H18.8583L23.4844 17.3427H31.4128ZM23.4844 19.6573L18.8583 27.7508H26.7775L31.4036 19.6573H23.4844Z" fill="#FF7500"></path></svg></a></div><div class="flex h-[34px] items-center justify-end justify-self-end"><button class="flex text-light-steel-gray text-sm items-center justify-center gap-x-[10px] rounded-[30px] border border-solid p-[10px] border-darker-steel-gray"><svg data-prefix="fas" data-icon="magnifying-glass" class="svg-inline--fa fa-magnifying-glass text-white" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M416 208c0 45.9-14.9 88.3-40 122.7L502.6 457.4c12.5 12.5 12.5 32.8 0 45.3s-32.8 12.5-45.3 0L330.7 376C296.3 401.1 253.9 416 208 416 93.1 416 0 322.9 0 208S93.1 0 208 0 416 93.1 416 208zM208 352a144 144 0 1 0 0-288 144 144 0 1 0 0 288z"></path></svg><span class="max-sm:hidden">⌘ <!-- -->+ K</span></button></div></div></div></header><main><div class="mx-auto max-w-(--breakpoint-xl) px-6"><ol class="block items-center gap-[8px] pt-[40px] uppercase"><li class="mr-[8px] inline-block"><a class="text-dark-steel-gray hover:text-white font-plexmono text-caption2" href="/library" data-discover="true">Library</a></li><li class="mr-[8px] inline-block"><svg width="5" height="8" viewBox="0 0 5 8" class="mt-[3px]" fill="none" xmlns="http://www.w3.org/2000/svg"><path d="M1 7L4 4L1 1" stroke="white" stroke-width="1.25" stroke-linecap="round" stroke-linejoin="round"></path></svg></li><li class="mr-[8px] inline-block"><a class="text-white font-plexmono text-caption2" href="/library/article/generative-ai-incident-response-devops" data-discover="true">Incident Response and DevOps in the Age of Generative AI</a></li><li class="mr-[8px] inline-block"></li></ol><div class="article text-body"><div class="bg-orange fixed bottom-0 left-0 z-20 h-[4px]" style="width:0%"></div><h1 class="font-planar text-title2 sm:text-title md:mt-[80px] md:mb-[100px] mb-4 md:text-center text-white"><span data-br=":R19ld5:" data-brr="1" style="display:inline-block;vertical-align:top;text-decoration:inherit;text-wrap:balance">Incident Response and DevOps in the Age of Generative AI</span><script>self.__wrap_n=self.__wrap_n||(self.CSS&&CSS.supports("text-wrap","balance")?1:2);self.__wrap_b=(n22, r2, e2) => {
e2 = e2 || document.querySelector(`[data-br="${n22}"]`);
let t2 = e2 == null ? void 0 : e2.parentElement;
if (!t2) return;
let l2 = (u2) => e2.style.maxWidth = u2 + "px";
e2.style.maxWidth = "";
let i2 = t2.clientWidth, d2 = t2.clientHeight, o2 = i2 / 2 - 0.25, s2 = i2 + 0.5, c22;
if (i2) {
for (l2(o2), o2 = Math.max(e2.scrollWidth, o2); o2 + 1 < s2; ) c22 = Math.round((o2 + s2) / 2), l2(c22), t2.clientHeight === d2 ? s2 = c22 : o2 = c22;
l2(s2 * r2 + i2 * (1 - r2));
}
e2.__wrap_o || (typeof ResizeObserver != "undefined" ? (e2.__wrap_o = new ResizeObserver(() => {
self.__wrap_b(0, +e2.dataset.brr, e2);
})).observe(t2) : process.env.NODE_ENV === "development" && console.warn("The browser you are using does not support the ResizeObserver API. Please consider add polyfill for this API to avoid potential layout shifts or upgrade your browser. Read more: https://github.com/shuding/react-wrap-balancer#browser-support-information"));
};self.__wrap_n!=1&&self.__wrap_b(":R19ld5:",1)</script></h1><div class="flex items-center justify-center"><ul class="flex flex-wrap gap-y-1 gap-x-2"><li class="inline-flex rounded-[20px] px-[13px] py-[8px] border border-dark-steel-gray"><a class="text-white text-label font-plexmono uppercase" href="/library?query=Artificial+Intelligence+(AI)" data-discover="true">Artificial Intelligence (AI)</a></li><li class="inline-flex rounded-[20px] px-[13px] py-[8px] border border-dark-steel-gray"><a class="text-white text-label font-plexmono uppercase" href="/library?query=Coding+Assistants" data-discover="true">Coding Assistants</a></li><li class="inline-flex rounded-[20px] px-[13px] py-[8px] border border-dark-steel-gray"><a class="text-white text-label font-plexmono uppercase" href="/library?query=LLM" data-discover="true">LLM</a></li><li class="inline-flex rounded-[20px] px-[13px] py-[8px] border border-dark-steel-gray"><a class="text-white text-label font-plexmono uppercase" href="/library?query=Incident+Response" data-discover="true">Incident Response</a></li><li class="inline-flex rounded-[20px] px-[13px] py-[8px] border border-dark-steel-gray"><a class="text-white text-label font-plexmono uppercase" href="/library?query=Generative+AI" data-discover="true">Generative AI</a></li><li class="inline-flex rounded-[20px] px-[13px] py-[8px] border border-dark-steel-gray"><a class="text-white text-label font-plexmono uppercase" href="/library?query=DevOps" data-discover="true">DevOps</a></li></ul></div><div class="bg-transparent -mx-6 px-6 rounded-[20px]"><div class="mx-[20px] mt-[50px] mb-[30px] h-px bg-darker-steel-gray"></div><div class="text-white text-caption2 uppercase font-plexmono mb-[30px] flex items-center justify-between md:px-[30px] flex-wrap gap-y-2"><div class="flex items-center gap-[8px]"><time dateTime="2023-10-12T22:08:00.000Z">Published <!-- -->OCT 12, 2023</time><div class="bg-darker-steel-gray h-[3px] w-[3px] rounded-full"></div><span>31<!-- --> mins</span></div><span class="flex items-center"><span class="mr-[12px] ml-auto block">Light Mode</span><label class="relative inline-block h-[34px] w-[60px] transition-all"><input type="checkbox" class="h-0 w-0 opacity-0"/><span class="bg-white/50 absolute top-0 right-0 bottom-0 left-0 cursor-pointer rounded-[34px] before:absolute before:bottom-[4px] before:left-[4px] before:h-[26px] before:w-[26px] before:rounded-[34px] before:bg-white before:transition-all before:content-[&#x27;&#x27;]"></span></label></span></div><div id="content" class="relative grid grid-cols-1 gap-y-[20px] lg:gap-x-[20px] lg:grid-cols-3"><div><div class=" "><div class="mt-[20px] lg:col-span-1"><ul class="border-dark-steel-gray mb-[20px] grid gap-[20px] rounded-[10px] border px-[20px] py-[20px]"><li><a href="/team/jesse-robbins" data-discover="true"><div class="flex w-full items-center gap-[10px]"><img class="h-[94px] w-[94px] shrink-0 rounded-full object-cover" src="https://cdn.sanity.io/images/50q6fr1p/production/9a8f35b91b4f52c518bf668debf266feeb8e2a4a-4000x6000.jpg?rect=1004,792,1981,1981&amp;w=94&amp;h=94&amp;auto=format&amp;dpr=2" alt="Jesse Robbins&#x27;s Headshot"/><div class="text-white flex flex-col gap-1"><span class="font-planar text-title5">Jesse Robbins</span><span class="text-caption2 font-plexmono block uppercase">General Partner<!-- -->, <span class="text-light-steel-gray">Heavybit</span></span></div></div></a></li></ul><div class="nested-scrollbar py-[20px] lg:overflow-y-auto lg:px-[30px] bg-light-steel-gray rounded-[10px] border-dark-steel-gray mt-[40px] border-l lg:border-none" style="max-height:calc(100vh - 20px)"><div class="max-w-xl pt-[20px]"><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">How Does Generative AI Work With Incident Response?</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">How to Utilize GenAI Within Incident Management Platforms</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-8"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Discussion: Where GenAI Makes Sense for IR with Nora Jones</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Where GenAI Impacts DevOps and the Future of Software Dev</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-8"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Discussion: What the Future Looks Like for Devs Using GenAI with Jeremy Edberg</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Deferring Low-Level Tasks to AI so Humans Can Focus on Strategy</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-8"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Discussion: The Division of Labor Between Human SRE and AI with Mandi Walls</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Why Humans in the Loop May Always Be Needed in SRE</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-8"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Discussion: The Brushes May Change, but Engineers Will Still Create Art with Brent Chapman</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">Conclusion</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div><div class="text-darker-steel-gray hover:text-heavy-slate text-caption2 font-plexmono mb-[20px] cursor-pointer ml-4 mb-[20px]"><div class="relative inline-block"><span class="mr-[10px] inline-block uppercase">More Resources:</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-heavy-slate absolute top-px -right-[15px] ml-[12px] -translate-x-full opacity-0 group-hover:opacity-100" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></div></div></div></div></div></div></div><div class="font-plexsans col-span-2 lg:ml-2 pt-[20px] pb-[10vh] text-light-steel-gray"><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="eb11bc467dae">How Does Generative AI Work With Incident Response?</h2><p>Software continues to eat the world, as more dev teams depend on third-party microservices as their daily infrastructure. Which means that outages are more common and costly than ever, costing upwards of <a rel="noopener noreferrer" href="https://llcbuddy.com/data/incident-management-statistics" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://llcbuddy.com/data/incident-management-statistics/">$100K</a> per incident, and that successful <a rel="noopener noreferrer" href="https://www.heavybit.com/devguild/incident-response" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.heavybit.com/devguild/incident-response">incident response workflows</a> are more important than ever as well. What about the wondrous wave of artificial intelligence products from Microsoft, GitHub, and OpenAI? Reports suggest generative AI tools boost <a rel="noopener noreferrer" href="https://arxiv.org/abs/2302.06590" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://arxiv.org/abs/2302.06590">developer productivity</a>, reducing bottlenecks by streamlining the process of coding with code snippets, search, and summaries. Could generative AI be a breakthrough for IT operations in managing incidents that helps stem this rising tide?</p><p><a rel="noopener noreferrer" href="https://www.axios.com/2023/08/09/ai-voters-trust-government-regulation" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.axios.com/2023/08/09/ai-voters-trust-government-regulation">62%</a> of the general populace is “concerned” about modern AI, and 86% “believe AI could accidentally cause a catastrophic event.” So where, if at all, does GenAI fit into incident reponse and day-to-day site reliability engineering? In this article, we consulted with a panel of site incident management veterans with more than 40 years of collective experience. As one of our experts put it, GenAI is good at “confidently delivering text that is pleasant to read, but not always complete, or correct.” As another suggested, GenAI is “not good at making decisions for you...or [emulating other people’s] expertise.”</p><p>Below, our panel explores known GenAI vulnerabilities in <a rel="noopener noreferrer" href="https://www.wired.com/story/generative-ai-prompt-injection-hacking" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.wired.com/story/generative-ai-prompt-injection-hacking/">security</a> and <a rel="noopener noreferrer" href="https://iapp.org/news/a/data-protection-issues-for-employers-to-consider-when-using-generative-ai" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://iapp.org/news/a/data-protection-issues-for-employers-to-consider-when-using-generative-ai/">privacy</a>, not to mention its well-documented <a rel="noopener noreferrer" href="https://www.wired.com/story/fast-forward-chatbot-hallucinations-are-poisoning-web-search" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.wired.com/story/fast-forward-chatbot-hallucinations-are-poisoning-web-search/">hallucinations</a>, and the need for ad hoc collaboration and consequential decisions in IM. Is there an eventual future for AI-powered incident commanders, or will teams always need that proverbial human in the loop? Our panel discusses:</p><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><strong class="text-body-bold text-white">The Strengths and Weaknesses of GenAI for IR and SRE:</strong> Which capabilities of GenAI are a strong fit for day-to-day incident management.</li><li><strong class="text-body-bold text-white">How GenAI Will Affect the DevOps and SRE Professions:</strong> How GenAI will impact professionals who work on both product and the operational side of product.</li><li><strong class="text-body-bold text-white">How GenAI Will Ultimately Affect Dev:</strong> Our panel also weighed in with their thoughts on how GenAI will impact the general business of software development.</li></ul><p><strong class="text-body-bold text-white"><em>Disclaimer: While the panelists interviewed here hail from companies such as Jeli, Amazon, and PagerDuty, the views expressed below are those of the individual panelists and do not reflect the views of their employers.</em></strong></p><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="217fdd6806f9">How to Utilize GenAI Within Incident Management Platforms</h2><div class="mt-[30px] mb-[40px] flex items-center justify-center"><img src="https://cdn.sanity.io/images/50q6fr1p/production/eeaae91fe0006fb9d98aced3b6ef71737f890e98-200x200.jpg?auto=format&amp;dpr=2" class="m:max-w-[500px]"/></div><p><a rel="noopener noreferrer" href="https://www.linkedin.com/in/norajones1" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.linkedin.com/in/norajones1/">Nora Jones</a> is an incident response veteran who led IR teams at Slack and Netflix before founding the developer-first IR startup <a rel="noopener noreferrer" href="https://www.jeli.io" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.jeli.io">Jeli</a>. She’s also a co-founder of the IR community <a rel="noopener noreferrer" href="https://www.learningfromincidents.io/about" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.learningfromincidents.io/about">LFI</a>. Her company has implemented <a rel="noopener noreferrer" href="https://techcrunch.com/2023/08/10/jeli-is-bringing-generative-ai-to-incident-report-analysis" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://techcrunch.com/2023/08/10/jeli-is-bringing-generative-ai-to-incident-report-analysis/">GenAI directly into its platform</a>, utilizing natural language to rapidly spin up shareable incident reports to quickly get team members up to speed as well as to draft overarching narratives based on different touch points over an incident’s life (including detection, diagnosis, and repair moments directly from chat logs, which the platform has already been fully annotating). Jones believes that GenAI’s ability to accelerate incident logging is a valuable tool and makes GenAI worth considering as another member of the team–but not as the overarching decision maker:</p><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><strong class="text-body-bold text-white">GenAI Has [At Least] Two Key Strengths for IR:</strong><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><em>Spinning Up Summaries to Catch Up Teams:</em> As incidents are happening, GenAI’s ability to quickly spin up content can be useful to provide instantaneous summaries of incidents to relevant team members to keep everyone in the loop as things happen (rather than pulling people sideways by requiring them to drop everything and hunt down the details).</li><li><em>Incident Analysis:</em> Post-incident, GenAI can help teams uncover and compile context around incidents to create richer, more-valuable post-mortems by collecting insights and notes across the incident lifecycle.</li></ul></li><li><strong class="text-body-bold text-white">Why GenAI May Not Be Taking the Incident Commander Chair Anytime Soon:</strong> Incident response continues to be a field full of unknowns and exceptions–not exactly a good fit for tools that are built largely to pattern-match based on past data. Fully AI-run incident remediation is unlikely to be “a thing” anytime soon.</li></ul><h3 class="text-lighter-steel-gray text-title4 mb-[10px]" id="1c027b6bf0c8">Discussion: Where GenAI Makes Sense for IR with Nora Jones</h3><p>In addition to offering the above observations, Jones opines that modern SRE can optimize their GenAI usage by recognizing its assorted strengths and weaknesses. Specifically, GenAI was not built on all-knowing, benevolent algorithms developed solely to decide how to manage important decisions, such as issue resolution steps. GenAI algorithms, at least for now, are optimized to generate and summarize content.</p><p>“I don&#x27;t think you trust GenAI to be an ‘expert’ in anything. It&#x27;s not good at making decisions for you. It&#x27;s not good at [emulating other people’s] expertise. It is good at summarizing pieces of information. But just because it&#x27;s easy to use and easy to ‘sprinkle’ AI on anything you&#x27;re doing doesn&#x27;t mean you should. I would really encourage folks that are starting to play around with it to understand actually how it works.”</p><blockquote class="my-10 ml-6 flex"><span class="mr-[15px] block"><svg data-prefix="fas" data-icon="quote-left" class="svg-inline--fa fa-quote-left text-orange text-2xl" role="img" viewBox="0 0 448 512" aria-hidden="true"><path fill="currentColor" d="M0 216C0 149.7 53.7 96 120 96l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64L0 216zm256 0c0-66.3 53.7-120 120-120l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64l0-136z"></path></svg></span><div class="text-white text-title4 font-planar">I think what we really want to do is use AI to get people more curious about what&#x27;s happening in their incidents.” -Nora Jones, Founder / Jeli</div></blockquote><p>“Ultimately, I think what we really want to do is use AI to get people more curious about what&#x27;s happening in their incidents. I’ve always believed that if you learn how an incident actually happens, you&#x27;ll be better off in the future. You can be more proactive about your incidents, resolving some of them more quickly, getting the right people in the room more quickly,” Jones explains. “Where AI seems really interesting for incident management is when we can use it to bubble up some of those interesting learnings, which then gets people investigating the incident...and gets people a little bit more curious about how it unfolded in the first place.”</p><p>Jones suggests that artificial intelligence provides opportunities to help both professional SREs and developers of all stripes. “I think GenAI will bring big changes in the field of incident management in terms of how incidents get communicated to stakeholders that are impacted by those incidents. But for developers in general, I think there’s an opportunity for them to use AI to accelerate their processes and help them get curious about other areas.” In the future, Jones suggests the possibility of AIs trained on large amounts of previous incident data being helpful in taking a more-proactive approach. “I don&#x27;t think generative AI is going to fix the incidents for you, but I think eventually, it might help point you to previous incidents that look like the one that you&#x27;re solving right now. But as far as I know, it can&#x27;t get people to talk to each other. And I don’t see it being a magic box for auto-remediation anytime soon.”</p><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="3a3e9cbcc658">Where GenAI Impacts DevOps and the Future of Software Dev</h2><div class="mt-[30px] mb-[40px] flex items-center justify-center"><img src="https://cdn.sanity.io/images/50q6fr1p/production/c5983aa5c027d4ed93b4206b4c981083c3fcfdc4-200x200.jpg?auto=format&amp;dpr=2" class="m:max-w-[500px]"/></div><p><a rel="noopener noreferrer" href="https://www.linkedin.com/in/jedberg" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.linkedin.com/in/jedberg/">Jeremy Edberg</a> is a longtime DevOps expert who currently helps lead Amazon’s Alexa Operational Excellence Team, but has done tours of duty at leading tech companies including eBay, Reddit, and Netflix–where he was a founding member of Netflix’s SRE team.</p><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><strong class="text-body-bold text-white">Large Language Models May Be the Next Major Evolutionary Step in Human-Computer Interfaces:</strong> Whether GenAI achieves the Nirvana-like goal of artificial general intelligence (AGI), prompt-based LLM chatbots may well represent the next step in the way humans interact with technology, as they effectively help computers take a big step toward being able to understand human language.</li><li><strong class="text-body-bold text-white">We’re Not Yet at a Point Where LLMs Can Credibly Recommend Remediation Steps:</strong> Right now, human users will trust their monitoring and alerting systems after those systems have proven themselves reliable, even going as far as allowing them to take automatic actions. But we are not there yet with LLMs. Right now, the best we can hope for is LLMs trained on previous incidents and patterns producing one (or a few) possible remediation steps and having a human select the best course of action. Over time, if the LLMs prove to produce the correct course of action in almost every case, they will be trusted to work autonomously.</li><li><strong class="text-body-bold text-white">Potential Career Evolution for DevOps:</strong> Language Model Operations in AIOps?: DevOps, being generally tasked with the maintenance and caretaking of infrastructure, may also inherit the care and feeding of language models. As LLMs come to represent more-significant components in infrastructure, organizations will need people who understand distributed computing, machine learning inference, managing GPUs and CPUs next to each other, storage, and other maintenance considerations. Could there be a point where entire careers are focused on monitoring AI models, updating them, and making sure the models are getting the right inputs and appropriately learning new things?</li></ul><h3 class="text-lighter-steel-gray text-title4 mb-[10px]" id="cd2a80857355">Discussion: What the Future Looks Like for Devs Using GenAI with Jeremy Edberg</h3><p>“Right now, GenAI is something of an advisory tool. We&#x27;re not to the point where we trust it enough to take the actions based on what it says,” Edberg explains. “In some ways, you could compare some of GenAI’s use cases to those of what monitoring used to be–or how things are when you&#x27;re first starting out because you don&#x27;t know that your monitoring and alerting are correct.”</p><p>“As an advisory tool, GenAI can tell you, ‘Hey, something is probably wrong here, and you should look into it,’ but there still needs to be a human in the loop there. Eventually, we&#x27;ll get to the point where we can take the human out of the loop for the easy stuff...maybe. The thing is, better monitoring and alerting have already made changes to the way we operate. And LLMs will definitely make changes to the way we operate, but there&#x27;ll be new challenges instead. Overall, I don’t think GenAI will eliminate DevOps jobs. It will, hopefully, make DevOps practices–and practitioners–much more efficient. So maybe in that regard, it would actually generate some net-new jobs.”</p><blockquote class="my-10 ml-6 flex"><span class="mr-[15px] block"><svg data-prefix="fas" data-icon="quote-left" class="svg-inline--fa fa-quote-left text-orange text-2xl" role="img" viewBox="0 0 448 512" aria-hidden="true"><path fill="currentColor" d="M0 216C0 149.7 53.7 96 120 96l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64L0 216zm256 0c0-66.3 53.7-120 120-120l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64l0-136z"></path></svg></span><div class="text-white text-title4 font-planar">In the future, if you are good at logic and want to learn how to reason about computer systems, [software engineering will still be] a great place to be.” -Jeremy Edberg, Principal Engineer / Amazon</div></blockquote><p>What effect will GenAI have on day-to-day dev workflows, or on software engineers as a profession? “If I were addressing a class of junior developers, I’d tell them, ‘GenAI is going to be a tool that will drastically speed up your development process, but it will not replace you.’ Not yet, anyway,” says Edberg. “Could it lower the barrier to entry for getting a job as an engineer? I could definitely see a situation where people–who hadn&#x27;t considered this type of career before, maybe because they weren&#x27;t interested in learning the details of coding syntax, for example, but are still good at general reasoning—might choose engineering now instead of business, law, or some other path.&quot;</p><p>&quot;Somebody who has these reasoning, logic, and analytical skills might be more interested now because the ‘hard parts’ are taken care of, the syntax, the math, that kind of stuff. In the future, I think if you are good at reasoning, good at logic, and want to learn how to reason about computer systems, it&#x27;s still a great place to be. If jobs do end up going away, they will be the ‘I&#x27;ve learned enough to know how to write code, and I&#x27;m going to spend most of my days writing basic, boilerplate&#x27; stuff,’ because the LLMs will take care of that.”</p><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="0b261b3fa906">Deferring Low-Level Tasks to AI so Humans Can Focus on Strategy</h2><div class="mt-[30px] mb-[40px] flex items-center justify-center"><img src="https://cdn.sanity.io/images/50q6fr1p/production/c87a2249f5a3a3696fc35defa73409803afeb2a6-200x200.jpg?auto=format&amp;dpr=2" class="m:max-w-[500px]"/></div><p><a rel="noopener noreferrer" href="https://www.linkedin.com/in/mandiwalls" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.linkedin.com/in/mandiwalls/">Mandi Walls</a> is a long-tenured developer advocate who has been advocating for AI and automated solutions to help make SRE and DevOps teams more productive for some time. She’s currently building communities of highly engaged developers at PagerDuty and has also served tours of duty at Chef and AOL.</p><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><strong class="text-body-bold text-white">We’re a Ways Off From Fully AI-Powered L1 Responders:</strong> The sheer amount of training datasets from real, recorded incidents across a single organization required to stand up completely AI-powered Level One incident responders just doesn’t exist yet. The closest path to something similar to this in the future might be from large orgs running similar services on a very similar platform with similar runtimes, which would presumably generate incidents of a similar character with similar symptoms.</li><li><strong class="text-body-bold text-white">The Most Immediate AI Opportunity in SRE May Be for Low-Level Remediation Tasks:</strong> Generative AI might be able to make the most immediate impact if it were trained to manage the low-level hiccups and false alarms that do not require extensive triage, and which experienced SREs resolve in minutes, anyway. There’s also strategic value in developing AI tools that can manage most low-level remediation tasks–because such tools would free up veteran SRE teams to focus more of their time and undivided attention on higher-priority incidents and post-mortems.</li></ul><h3 class="text-lighter-steel-gray text-title4 mb-[10px]" id="9e82dab8ea81">Discussion: The Division of Labor Between Human SRE and AI with Mandi Walls</h3><p>Walls suggests that the immediate value of generative AI in SRE might come from spinning up documentation and after-action reports, but also in a variety of other areas. “Our incident response process includes Zoom calls, recordings, transcripts, and Slack channels, along with charts and graphs and many other kinds of data and artifacts...it’s a slog. So there’s value in letting AI generate all the components and artifacts we need.”</p><p>Regarding how GenAI and its associated tools, such as code generators, could affect the profession of development as a whole, Walls sees opportunities in many areas for GenAI to be valuable. “Stuff like coding assistants are super interesting. Some of it is really clever and is already doing a really good job for folks doing some of that work. But as someone who uses a lot of products, I&#x27;m hoping for improved documentation—and API documentation in particular–that developers don&#x27;t have to write themselves. It’d be good to see tools improve enough to automatically generate all that stuff and make it more useful.”</p><p>Walls suggests that testing may be another area of opportunity for GenAI to improve development pipelines. “Another use case would be generating tests. I think there&#x27;s a lot of knowledge already in that space, especially over the last 10 years as that whole practice has become more automated, and maybe it will become even more so. So maybe the work of test engineers will move more towards creating better harnesses and doing performance monitoring on the testing process rather than anything like writing a tool. Also, developers have artifact repositories. There&#x27;s all this stuff that has to run together really closely. And keeping that all in line plus maintaining changes that come in from the vendors would definitely be helped by additional tooling that&#x27;s a little bit smarter than what we have right now.”</p><blockquote class="my-10 ml-6 flex"><span class="mr-[15px] block"><svg data-prefix="fas" data-icon="quote-left" class="svg-inline--fa fa-quote-left text-orange text-2xl" role="img" viewBox="0 0 448 512" aria-hidden="true"><path fill="currentColor" d="M0 216C0 149.7 53.7 96 120 96l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64L0 216zm256 0c0-66.3 53.7-120 120-120l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64l0-136z"></path></svg></span><div class="text-white text-title4 font-planar">For positions like SRE that are usually more directly integrated with an engineering practice, they&#x27;ll see more benefits from coding tools, which could start to learn as much about infrastructure tooling as they do about regular languages and runtime-application code.” -Mandi Walls, Developer Advocate / PagerDuty</div></blockquote><p>“I’m thinking about something that’s even a level up from Dependabot–which right now, will send you an email that says, ‘Hey, here&#x27;s this thing that needs to be updated.’ It would be really useful to see this kind of use case broadening out to alert you that your vendor is doing an upgrade. An alert that could tell you, ‘Here&#x27;s what we recommend for your specific use case.’ For example, if your cloud provider is turning off instances of your level, here&#x27;s where you need to migrate...and then starting to do that work for you without having to really intervene.”</p><p>On how GenAI may affect the business of DevOps workflows, Walls is less eager to make predictions due to variance across orgs. “DevOps jobs are different in every organization. So it&#x27;s possible that GenAI tools, such as code generators, could make a difference because, in some places, a lot of those folks are writing more code. However, in other places, they&#x27;re just working more in advisory positions. And then, some orgs take more of a build-and-release approach. So it&#x27;s hard to stay. At the macro level, if there&#x27;s going to be a deep change in what it means to work in DevOps due to generative AI, well...I&#x27;m not sure there&#x27;s enough of a consensus of what a DevOps engineer should be doing to be able to say that.”</p><p>On how GenAI will affect the business of SRE, Walls is significantly more bullish. “I think for those positions like SRE that are usually more directly integrated with an engineering practice, I think they&#x27;ll see more benefits from coding tools and things like that...which potentially could start to learn as much about infrastructure tooling as they do about regular languages and runtime–application code versus infrastructure code. I’d like to see GenAI help SRE teams push forward along their golden path because so much of their infrastructure is hopefully managed as code.” And in the same way that developer technology such as containers expanded into open source with Kubernetes, there may be opportunities to see open source contribute to generative coding assistants for SREs. “I think these teams will also benefit from those same code generation tools–but they may be in Terraform or Pulumi, rather than Python/Elixir/Go/Rust.”</p><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="46adbb984ffc">Why Humans in the Loop May Always Be Needed in SRE</h2><div class="mt-[30px] mb-[40px] flex items-center justify-center"><img src="https://cdn.sanity.io/images/50q6fr1p/production/28240d630e8fb1c07da810841e34f41fd91de686-200x200.jpg?auto=format&amp;dpr=2" class="m:max-w-[500px]"/></div><p><a rel="noopener noreferrer" href="https://www.linkedin.com/in/brentchapman" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.linkedin.com/in/brentchapman/">Brent Chapman</a> is a pioneer in what is now known as modern SRE. Throughout his career in technology, he has always also worked as a volunteer in public safety and emergency services, starting as a search-and-rescue pilot and incident commander for air search and rescue. He applied the principles he learned in emergency services to his tenure at Google, where developed the company’s internal <a rel="noopener noreferrer" href="https://sre.google/workbook/incident-response" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://sre.google/workbook/incident-response/">Incident Management at Google (IMAG)</a> practice, and later, brought similar foundational practices to Slack. He currently runs the incident management consultancy <a rel="noopener noreferrer" href="https://greatcircle.com" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://greatcircle.com/">Great Circle Associates</a>.</p><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><strong class="text-body-bold text-white">Large Language Models Have the “Natural Language” Part Down, But May Still Lack in Other Areas:</strong> While things may certainly change in the future, LLM chatbots seem best at confidently delivering text that is pleasant to read, but not always complete, or correct, and certainly not above human verification. Today’s LLMs are sometimes wrong, but never uncertain.</li><li><strong class="text-body-bold text-white">GenAI’s Greatest Value to SRE Might Be for After Reports:</strong> Post-incident phases call for a great many write-ups to document the conditions leading up to outages, the circumstances and effects of the outages, and the actions taken to resolve the outages. GenAI can certainly produce something readable and user-friendly for general audiences, though expert engineers may prefer to keep all the gory details.</li><li><strong class="text-body-bold text-white">Maybe There’s a Future for AI-Powered Pattern-Matching and Timeframe Planning in IM:</strong> Chapman recalls his years working with highly experienced engineers who were so well-versed in their systems that they seemed to have a sixth sense when it came to browsing a series of graphs and detecting seemingly imperceptible irregularities when investigating root causes. Could AI tools eventually become “smart” enough to detect such inconsistencies? Maybe. They might be even more useful in helping size incidents and projected response time windows that correspond with severity.</li></ul><h3 class="text-lighter-steel-gray text-title4 mb-[10px]" id="ba5d422e08c3">Discussion: The Brushes May Change, but Engineers Will Still Create Art with Brent Chapman</h3><p>As Chapman reflects on the fundamental practice of incident management, he finds few intersections with GenAI’s biggest strengths and sees many of the processes as still being fundamentally human. “In incident management, the challenge is always that we know something has gone wrong, but we don&#x27;t always know what has gone wrong, or who needs to do what to fix it. The process is about figuring out the details, executing that response, getting things back to a stable situation, and then getting fully recovered to a normal state of operation. All very challenging activities that involve working under time pressure that people normally don&#x27;t have, working across teams that don&#x27;t routinely work together and don&#x27;t know each other&#x27;s capabilities and concerns and considerations and so forth.”</p><p>“In our day-to-day work, we establish project teams. We spend a lot of time ‘storming and norming’ to build that whole framework of ‘how do we learn about each other and work together effectively,’ and have debates and arguments and joint planning activities that let technology companies do the amazing things they do,” Chapman offers. But in the same way that Agile methodology teams emphasize flexibility and a pragmatic approach to delivery, incident teams also need to be realistic.”</p><p>“Those things all take time and energy to establish. And you don&#x27;t have that time and energy available during an emergency.” Downtime doesn’t just provide opportunities for collaboration–it demands collaboration. “You need to find a way to work together quickly and effectively enough for the emergency, even if it&#x27;s not necessarily a great way to work together in the long run. It&#x27;s very top-down, it&#x27;s very authoritative, it&#x27;s very hierarchical, it&#x27;s very old-fashioned. But it works better in an emergency. Also, not everybody who&#x27;s going to help you will be available at the same time, and certainly not at the start of the incident. You have to start responding with who&#x27;s available at the time and incorporate more people over time. You need to have ways of effectively putting people to work and then putting more people in the process without disrupting the work that&#x27;s already in progress, so people can come up to speed without disrupting those activities and then plug themselves in, offload, or take on some new tasks related to the emergency. And somebody has to manage and coordinate all of this and manage the communications.”</p><blockquote class="my-10 ml-6 flex"><span class="mr-[15px] block"><svg data-prefix="fas" data-icon="quote-left" class="svg-inline--fa fa-quote-left text-orange text-2xl" role="img" viewBox="0 0 448 512" aria-hidden="true"><path fill="currentColor" d="M0 216C0 149.7 53.7 96 120 96l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64L0 216zm256 0c0-66.3 53.7-120 120-120l8 0c17.7 0 32 14.3 32 32s-14.3 32-32 32l-8 0c-30.9 0-56 25.1-56 56l0 8 64 0c35.3 0 64 28.7 64 64l0 64c0 35.3-28.7 64-64 64l-64 0c-35.3 0-64-28.7-64-64l0-136z"></path></svg></span><div class="text-white text-title4 font-planar">[Working with GenAI is] a lot like dealing with a very junior programmer. You still have to check it. You still have to write the code. You still have to write the test.” -Brent Chapman, Principal / Great Circle Associates</div></blockquote><p>Chapman suggests that part of the excitement, and confusion, around AI and its benefits may come from an excessive widening or narrowing of definitions. The SRE veteran discusses automated systems that he worked on at Google for real-time ‘traffic’ (server cluster load balancing) management. “I worked on a system that tracked incident response patterns whenever there were problems with a given cluster, such that the first thing we’d do is drain incoming traffic away from that cluster, and send the traffic somewhere else that&#x27;s still healthy. Realizing that this was the first thing we almost always did, we decided to automate that. But we needed some guardrails. For instance, we didn&#x27;t want to drain traffic away from the last cluster in any continent, or from a cluster when there were only two clusters left. So I built this system in Python which, when it was alerted to unhealthy clusters, it would ‘think’ about draining it, run through the list of checks for that service, and decide whether or not to drain the service in that cluster. All taking place while people were still responding to their pagers. But depending on your definitions, this might be more of a style of ‘mechanical turk’-style automation rather than ‘AI,’ I suppose.”</p><p>What will the future hold for software engineers? For operations, Chapman sees potential in highly-trained AI systems with the ability to take certain actions, such as taking steps to provision new systems, or adjusting configurations, autonomously. “For instance, if you tie a generative AI system to your AWS console–your control system for your cloud computing system–and you can start having a conversation with it about, let’s say, bringing online another 20% of capacity in London. And the system replies that there&#x27;s not enough spare capacity in London, but it can give you 15% in London and 5% across Europe, for example. You can imagine starting to have these sorts of operative discussions with it that are going to result in things happening...under approval with your supervision, and so forth. Right now, a lot of our monitoring control systems present human operators with problems, and it&#x27;s up to the operator to solve them. I think the next step is going to be to have generative AI propose solutions. Now, is it going to propose better solutions than your average new hire six months out of college? Maybe. That&#x27;s going to be an interesting question.”</p><p>On the topic of AI’s impact on developers as a whole, Chapman still feels strongly that systems will ultimately still need humans in the loop. “I think there&#x27;s going to continue to be a need for more developers. However, there&#x27;s going to be a new skill set that many of them have: ‘prompting,’ basically developing prompts, asking the right questions, feeding the right data to get a useful result out of the generative AI systems they&#x27;re working with. It seems comparable to art. ‘Creating art’ is going to change from knowing how to mix paints and pick a brush and applying certain physical techniques to knowing how to describe what you&#x27;re looking for to the AI...so that it can generate something that looks like what you want. I&#x27;m already hearing a bunch of my programmer colleagues talking about how they’ve had good luck using ChatGPT for tasks such as providing the framework of a Python application that does such and such. It’ll write the first hundred lines of code and create the first 10 files, and basically sets up your project for you. And then you can go from there. But it&#x27;s a lot like dealing with a very junior programmer who doesn&#x27;t always understand your intent and sometimes just goes off into the weeds. You still have to check it. You still have to write the code. You still have to write the test. I’ve also heard some people report success generating unit tests based on inputted code (which seems kind of backward, since you’re supposed to write your tests first and write the code to make the test pass), but is that better than no unit tests at all? Yeah, probably.”</p><p>“One of my favorite science fiction authors is <a rel="noopener noreferrer" href="https://en.wikipedia.org/wiki/Vernor_Vinge" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://en.wikipedia.org/wiki/Vernor_Vinge">Vernor Vinge</a>, who was a professor of computer science at San Diego State University. He wrote a story with a character who, for various reasons having to do with relativistic time dilation and traveling near the speed of light, ends up returning home a thousand years later, despite only aging about 10 years. He believes his skills are going to be completely out of date in this new world. And it turns out, the guy&#x27;s a programmer and there is a place for his skills as basically someone who understands the system 14 layers underneath what the AIs are doing in that present day. He has an ability, by understanding those underpinnings, to bypass a lot of layers and go straight to the ‘low level,’ a bit like a programmer today who still understands assembly language. Obviously, this is kind of a simplified example, but I think there are still going to be plenty of roles for humans in technology. Someone still has to verify–to ask the questions: ‘Is this right? Is this useful? Is this complete?’ I don&#x27;t see that role moving away from human judgment.”</p><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="8c3185928299">Conclusion</h2><p>GenAI is already seeing direct applications in day-to-day SRE work. For more information and discussion on how AI may affect the future of incident management, DevOps, and the business of software development as a whole, join the <a rel="noopener noreferrer" href="https://www.heavybit.com/devguild" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.heavybit.com/devguild">DevGuild: AI Summit</a> event.</p><h2 class="text-lighter-steel-gray text-title2 md:text-title3 xl:text-title3 mb-[10px]" id="701eb23619a8">More Resources:</h2><ul class="mx-[40px] mb-[20px] flex list-disc flex-col gap-[5px]"><li><a rel="noopener noreferrer" href="https://www.heavybit.com/library/article/incident-response-best-practices" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.heavybit.com/library/article/incident-response-best-practices">Article - Three Key Best Practices for Modern Incident Response</a></li><li><a rel="noopener noreferrer" href="https://www.heavybit.com/devguild/incident-response" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.heavybit.com/devguild/incident-response">On-Demand Video Series - DevGuild: Incident Response</a></li><li><a rel="noopener noreferrer" href="https://www.heavybit.com/library/podcasts/getting-there/ep-7-the-march-2023-datadog-outage-with-laura-de-vesine" target="_blank" class="hover:text-orange decoration-orange underline decoration-1 underline-offset-[3px]" title="Navigate to https://www.heavybit.com/library/podcasts/getting-there/ep-7-the-march-2023-datadog-outage-with-laura-de-vesine">Podcast - Getting There with Nora Jones, Niall Murphy and Laura De Vesine</a></li></ul></div></div></div><div class="mx-auto max-w-(--breakpoint-lg)"><div class="mt-[100px]"><section class="py-[50px]"><div class="mb-[20px] flex flex-col justify-between gap-[20px] sm:flex-row sm:items-center"><h3 class="font-plexmono text-caption2 uppercase text-white">Content from the Library</h3><div><a class="text-caption2 font-plexmono inline-flex items-center justify-center gap-[10px] text-white uppercase" href="/library?query=" data-discover="true"><span>Visit library</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-white" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></a></div></div><div class="relative grid grid-cols-1 gap-[20px] md:grid-cols-3"><a class="group border-dark-steel-gray row-span-2 block rounded-[10px] border-[4px] border-solid bg-white" href="/library/collections/artificial-intelligence-for-startup-founders" data-discover="true"><div class="flex h-full flex-col justify-between px-[20px] pt-[20px] pb-[30px]"><div class="h-[257px] w-full rounded-[10px]" style="background-color:[object Object]"><div class="text-title4 font-planar py-[29px] text-black">AI for Startup Founders</div></div><div class="font-planar text-body2 text-darker-steel-gray my-5">This article is part of the &#x27;AI for Startup Founders&#x27; Collection Series</div><div class="flex justify-end"><button class="group-hover:bg-heavy-slate border-heavy-slate flex items-center justify-center gap-[10px] rounded-[20px] border border-solid px-[20px] py-[12px] transition-all"><span class="text-heavy-slate text-caption2 font-plexmono uppercase transition-all group-hover:text-white">Go to collection</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-orange" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></button></div></div></a><a class="row-span-2 group block rounded-[10px] bg-white" href="/library/video/open-source-licensing-and-the-future-of-open-source-businesses" data-discover="true"><div class="flex flex-col justify-between px-[20px] pt-[20px] pb-[30px] h-full"><div><div class="flex items-center gap-[8px]"><time class="text-caption2 text-darker-steel-gray font-plexmono uppercase">Jun 29, 2023</time><div class="bg-darker-steel-gray h-[3px] w-[3px] rounded-full"></div><small class="text-caption2 text-darker-steel-gray font-plexmono uppercase">Video</small></div><h3 class="font-planar text-title4 mt-[20px] text-black">Open-Source Licensing and The Future of Open Source Businesses</h3><p class="text-darker-steel-gray text-body2 font-planar mt-[10px]">In this panel we&#x27;ll dig into the current state of open source licensing and what is and is not open source, what the current...</p></div><div class="mt-[40px] flex justify-end"><button class="group-hover:bg-heavy-slate border-heavy-slate flex items-center justify-center gap-[10px] rounded-[20px] border border-solid px-[20px] py-[12px] transition-all"><svg data-prefix="fas" data-icon="play" class="svg-inline--fa fa-play text-orange" role="img" viewBox="0 0 448 512" aria-hidden="true"><path fill="currentColor" d="M91.2 36.9c-12.4-6.8-27.4-6.5-39.6 .7S32 57.9 32 72l0 368c0 14.1 7.5 27.2 19.6 34.4s27.2 7.5 39.6 .7l336-184c12.8-7 20.8-20.5 20.8-35.1s-8-28.1-20.8-35.1l-336-184z"></path></svg><span class="text-heavy-slate text-caption2 font-plexmono uppercase transition-all group-hover:text-white">Watch</span><div class="bg-heavy-slate h-[3px] w-[3px] rounded-full transition-all group-hover:bg-white"></div><span class="text-heavy-slate text-caption2 font-plexmono uppercase transition-all group-hover:text-white">31<!-- --> <!-- -->MIN</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-orange" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></button></div></div></a><a class="row-span-2 group block rounded-[10px] bg-white" href="/library/article/why-on-device-inference-needs-custom-observability" data-discover="true"><div class="flex flex-col justify-between px-[20px] pt-[20px] pb-[30px] h-full"><div><div class="flex items-center gap-[8px]"><time class="text-caption2 text-darker-steel-gray font-plexmono uppercase">Aug 25, 2026</time><div class="bg-darker-steel-gray h-[3px] w-[3px] rounded-full"></div><small class="text-caption2 text-darker-steel-gray font-plexmono uppercase">Article</small></div><h3 class="font-planar text-title4 mt-[20px] text-black">Why On-Device Inference Needs Custom Observability</h3><p class="text-darker-steel-gray text-body2 font-planar mt-[10px]">The Unique Challenges of Mobile Compute
A significant focus in modern AI has been on large language models with billions of...</p></div><div class="mt-[40px] flex justify-end"><button class="group-hover:bg-heavy-slate border-heavy-slate flex items-center justify-center gap-[10px] rounded-[20px] border border-solid px-[20px] py-[12px] transition-all"><svg data-prefix="fas" data-icon="bookmark" class="svg-inline--fa fa-bookmark text-orange" role="img" viewBox="0 0 384 512" aria-hidden="true"><path fill="currentColor" d="M64 0C28.7 0 0 28.7 0 64L0 480c0 11.5 6.2 22.2 16.2 27.8s22.3 5.5 32.2-.4L192 421.3 335.5 507.4c9.9 5.9 22.2 6.1 32.2 .4S384 491.5 384 480l0-416c0-35.3-28.7-64-64-64L64 0z"></path></svg><span class="text-heavy-slate text-caption2 font-plexmono uppercase transition-all group-hover:text-white">Read</span><div class="bg-heavy-slate h-[3px] w-[3px] rounded-full transition-all group-hover:bg-white"></div><span class="text-heavy-slate text-caption2 font-plexmono uppercase transition-all group-hover:text-white">13<!-- --> <!-- -->MIN</span><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-orange" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></button></div></div></a></div></section></div></div></div></div></main><footer class="bg-orange mt-[50px] pt-[50px] pb-[20px]"><div class="mx-auto max-w-(--breakpoint-xl) px-6"><div class="mb-[20px] flex flex-col justify-between gap-x-[50px] lg:flex-row"><div class="basis-1/3"><h3 class="text-title3 font-planar mb-[26px] text-white"><svg width="225" height="43" viewBox="0 0 225 43" fill="none" xmlns="http://www.w3.org/2000/svg"><path fill-rule="evenodd" clip-rule="evenodd" d="M0.397949 21.078L11.9109 41.5864H34.9035L46.4164 21.078L34.9035 0.56958H11.9109L0.397949 21.078ZM37.7983 21.078L30.6111 33.8915H16.2366L9.04929 21.078L16.2366 8.26446H30.6111L37.7983 21.078Z" fill="white"></path><path fill-rule="evenodd" clip-rule="evenodd" d="M34.2048 19.7894L29.1471 10.8403H20.5291L25.5535 19.8233H34.2048V19.7894ZM25.5535 22.3657L20.5291 31.3487H29.1471L34.1715 22.3657H25.5535Z" fill="white"></path><path d="M96.0952 36.6718C101.02 36.6718 105.046 33.5192 106.244 29.0447H100.754C100.254 29.8243 99.5557 30.4345 98.7238 30.8752C97.9253 31.3158 96.9936 31.5192 96.0952 31.4853C93.0006 31.4853 90.6382 29.282 90.6382 26.2311V24.4345C90.6382 21.1125 93.0006 18.7396 96.0952 18.7396C99.1897 18.7396 101.486 21.1125 101.486 24.4345L102.317 23.2481H89.8396L89.9727 27.1125H106.51C106.577 26.5023 106.61 25.8921 106.61 25.282C106.61 18.9091 101.985 13.96 96.0952 13.96C90.2056 13.96 85.5472 18.9091 85.5472 25.282C85.5472 31.7226 90.1723 36.7057 96.0952 36.6718ZM207.997 19.6888H213.753V36.1294H225V30.9769H218.878V19.6888H225V14.5363H218.878V5.75664H213.753V14.5363H207.997V19.6888ZM200.81 36.1294H205.967V14.5363H200.81V36.1294ZM200.078 7.72273C200.078 8.1634 200.144 8.63798 200.311 9.04476C200.477 9.48543 200.71 9.85831 201.009 10.1973C201.309 10.5363 201.708 10.7736 202.107 10.977C202.507 11.1464 202.972 11.2481 203.405 11.2481C203.838 11.2481 204.303 11.1464 204.703 10.977C205.102 10.8075 205.501 10.5363 205.801 10.1973C206.1 9.85831 206.366 9.48543 206.5 9.04476C206.666 8.60408 206.732 8.1634 206.732 7.72273C206.699 6.84138 206.333 6.02782 205.701 5.41766C205.069 4.80749 204.27 4.46851 203.405 4.46851C202.54 4.46851 201.708 4.80749 201.109 5.41766C200.477 6.02782 200.111 6.87527 200.078 7.72273ZM181.61 25.3498C181.61 21.8244 184.073 19.2142 187.4 19.2142C190.727 19.2142 193.19 21.7905 193.19 25.3498C193.19 28.8752 190.727 31.4514 187.4 31.4514C184.073 31.4514 181.61 28.8413 181.61 25.3498ZM176.586 36.1294H181.377V33.9938C182.243 34.8413 183.241 35.5192 184.339 35.9938C185.437 36.4684 186.635 36.6718 187.833 36.6718C193.689 36.6718 198.281 31.7226 198.281 25.3498C198.281 18.9769 193.689 13.9939 187.833 13.9939C185.57 13.9939 183.407 14.8074 181.71 16.3329V4.638H176.586V36.1294ZM153.893 14.5363L162.511 33.8582L159.416 42.57H164.674L175.288 14.5363H169.964L165.04 27.5871L159.45 14.5363H153.893ZM131.333 14.5363L139.951 36.1294H145.008L153.627 14.5363H148.303L142.546 29.2142L136.79 14.5363H131.333ZM113.531 25.3498C113.531 21.8244 116.027 19.2142 119.321 19.2142C122.615 19.2142 125.144 21.7905 125.144 25.3498C125.144 28.8752 122.648 31.4514 119.321 31.4514C115.993 31.4514 113.531 28.8413 113.531 25.3498ZM108.407 25.3498C108.407 31.7226 112.999 36.6718 118.888 36.6718C121.251 36.6718 123.547 35.7565 125.277 34.0955V36.1294H130.035V14.5363H125.277V16.604C123.547 14.9091 121.284 13.9939 118.888 13.9939C112.999 13.96 108.407 18.943 108.407 25.3498ZM57.9294 36.1294H63.3199V23.1464H77.7277V36.1294H83.1182V4.6041H77.7277V17.621H63.3199V4.6041H57.9294V36.1294Z" fill="white"></path></svg></h3><div class="text-body font-plexsans mb-[121px] text-white">Unlock success from technical pioneers in software infrastructure. We send the best strategies, tools, and industry updates directly to your inbox.</div><form action="https://share.hsforms.com/1_KYaR4lAQm-C8_9H7r4MIwqdoyu" method="GET" target="_blank"><div class="relative"><input type="hidden" name="utm_source" value="/library/article/generative-ai-incident-response-devops"/><input type="email" id="email" name="email" required="" placeholder="ENTER EMAIL" class="sm:placeholder:text-caption2 placeholder:text-body2 placeholder:font-planar font-planar sm:text-caption2 text-body2 block w-full border-b border-solid border-white bg-transparent px-2 py-8 text-white outline-hidden placeholder:text-white"/><button type="submit" class="absolute top-1/2 right-0 flex h-[34px] w-[34px] -translate-y-1/2 items-center justify-center rounded-full bg-white"><svg data-prefix="fas" data-icon="arrow-right" class="svg-inline--fa fa-arrow-right text-orange" role="img" viewBox="0 0 512 512" aria-hidden="true"><path fill="currentColor" d="M502.6 278.6c12.5-12.5 12.5-32.8 0-45.3l-160-160c-12.5-12.5-32.8-12.5-45.3 0s-12.5 32.8 0 45.3L402.7 224 32 224c-17.7 0-32 14.3-32 32s14.3 32 32 32l370.7 0-105.4 105.4c-12.5 12.5-12.5 32.8 0 45.3s32.8 12.5 45.3 0l160-160z"></path></svg></button></div></form></div><div class="mt-[80px] grid max-w-(--breakpoint-sm) basis-2/3 grid-cols-2 gap-[50px] sm:flex sm:justify-between sm:gap-0 lg:mt-0"><ul class="flex flex-col gap-y-[20px]"><li><a href="/portfolio" class="font-plexmono text-caption2 text-white uppercase" data-discover="true">Portfolio</a></li><li><a href="/team" class="font-plexmono text-caption2 text-white uppercase" data-discover="true">Team</a></li><li><a href="/press" class="font-plexmono text-caption2 text-white uppercase" data-discover="true">News</a></li></ul><ul class="flex flex-col gap-y-[20px]"><li><a href="/events" class="font-plexmono text-caption2 text-white uppercase" data-discover="true">Events</a></li><li><a rel="noopener noreferrer" href="https://www.heavybit.com/devguild" class="font-plexmono text-caption2 text-white uppercase" target="_blank">DevGuild</a></li><li><a href="/jobs" class="font-plexmono text-caption2 text-white uppercase" data-discover="true">Jobs</a></li><li><a href="/community/diversity-and-inclusion" class="font-plexmono text-caption2 text-white uppercase" data-discover="true">Inclusion</a></li></ul><ul class="flex flex-col gap-y-[20px]"><li><a rel="noopener noreferrer" href="https://www.heavybit.com/library" class="font-plexmono text-caption2 text-white uppercase" target="_blank">Blog</a></li><li><a rel="noopener noreferrer" href="https://www.heavybit.com/library/podcasts" class="font-plexmono text-caption2 text-white uppercase" target="_blank">Podcast Network</a></li><li><a rel="noopener noreferrer" href="https://www.heavybit.com/devtoolsdigest" class="font-plexmono text-caption2 text-white uppercase" target="_blank">DevToolsDigest</a></li></ul></div></div><ul class="mb-[24px] ml-auto flex w-8 items-center gap-x-[24px] sm:justify-end"><li class="h-6 w-6"><a target="_blank" href="https://twitter.com/heavybit"><img src="https://cdn.sanity.io/images/50q6fr1p/production/9d618da0948299428b29e5d499af3fa5369a24d2-20x20.svg?auto=format&amp;dpr=2"/></a></li><li class="h-6 w-6"><a target="_blank" href="https://www.linkedin.com/company/heavybit"><img src="https://cdn.sanity.io/images/50q6fr1p/production/01bd92eb06873bdc30b1cb5f9c78b1ef5a9559d4-512x512.png?auto=format&amp;dpr=2"/></a></li><li class="h-6 w-6"><a target="_blank" href="https://www.youtube.com/c/heavybit"><img src="https://cdn.sanity.io/images/50q6fr1p/production/6c548d2e815d1fb3b3077c9b8be1ec251e1bc418-512x512.png?auto=format&amp;dpr=2"/></a></li><li class="h-6 w-6"><a target="_blank" href="https://www.instagram.com/heavybitindustries/"><img src="https://cdn.sanity.io/images/50q6fr1p/production/b38b7ee9ccd05f61ca43d5b29fe7058f822f87eb-512x512.webp?auto=format&amp;dpr=2"/></a></li><li class="h-6 w-6"><a target="_blank" href="https://www.tiktok.com/@heavybitindustries"><img src="https://cdn.sanity.io/images/50q6fr1p/production/48ac9473e30341c4bee9f6634b77b6f8f1dd6b2f-849x980.webp?auto=format&amp;dpr=2"/></a></li></ul><div class="items-center justify-between border-t border-solid border-white/30 pt-[30px] sm:flex"><p class="text-caption2 font-plexmono text-white uppercase">Heavybit is the leading fund for developer and enterprise startups.</p><span class="text-caption2 font-plexsans mt-[10px] block text-white uppercase sm:mt-0">© <!-- -->2026<!-- --> Heavybit</span></div></div></footer><script>((storageKey2, restoreKey) => {
if (!window.history.state || !window.history.state.key) {
let key = Math.random().toString(32).slice(2);
window.history.replaceState({ key }, "");
}
try {
let positions = JSON.parse(sessionStorage.getItem(storageKey2) || "{}");
let storedY = positions[restoreKey || window.history.state.key];
if (typeof storedY === "number") {
window.scrollTo(0, storedY);
}
} catch (error) {
console.error(error);
sessionStorage.removeItem(storageKey2);
}
})("react-router-scroll-positions", null)</script><link rel="modulepreload" href="/assets/entry.client-yKU2CyKw.js"/><link rel="modulepreload" href="/assets/chunk-QMGIS6GS-BqlQPwNs.js"/><link rel="modulepreload" href="/assets/index-DCUuxq93.js"/><link rel="modulepreload" href="/assets/root-CJcvKFFi.js"/><link rel="modulepreload" href="/assets/urlFor-BU3g6tJ3.js"/><link rel="modulepreload" href="/assets/index-LUh8MTnN.js"/><link rel="modulepreload" href="/assets/library.article._slug-C6Exb9Ab.js"/><link rel="modulepreload" href="/assets/index-DTAFgDfc.js"/><link rel="modulepreload" href="/assets/blogArticle-CdYFL8Rt.js"/><link rel="modulepreload" href="/assets/index-2PDpsLtL.js"/><link rel="modulepreload" href="/assets/page-DKcPV4FZ.js"/><link rel="modulepreload" href="/assets/index-DY5Afvo1.js"/><link rel="modulepreload" href="/assets/Heavybit-CW9lClJ7.js"/><link rel="modulepreload" href="/assets/index-Bm3smRr1.js"/><script>window.__reactRouterContext = {"basename":"/","future":{"unstable_middleware":false,"unstable_optimizeDeps":false,"unstable_splitRouteModules":false,"unstable_subResourceIntegrity":false,"unstable_viteEnvironmentApi":false},"routeDiscovery":{"mode":"lazy","manifestPath":"/__manifest"},"ssr":true,"isSpaMode":false};window.__reactRouterContext.stream = new ReadableStream({start(controller){window.__reactRouterContext.streamController = controller;}}).pipeThrough(new TextEncoderStream());</script><script type="module" async="">;
import * as route0 from "/assets/root-CJcvKFFi.js";
import * as route1 from "/assets/library.article._slug-C6Exb9Ab.js";
window.__reactRouterManifest = {
"entry": {
"module": "/assets/entry.client-yKU2CyKw.js",
"imports": [
"/assets/chunk-QMGIS6GS-BqlQPwNs.js",
"/assets/index-DCUuxq93.js"
],
"css": []
},
"routes": {
"root": {
"id": "root",
"path": "",
"hasAction": false,
"hasLoader": true,
"hasClientAction": false,
"hasClientLoader": false,
"hasClientMiddleware": false,
"hasErrorBoundary": true,
"module": "/assets/root-CJcvKFFi.js",
"imports": [
"/assets/chunk-QMGIS6GS-BqlQPwNs.js",
"/assets/index-DCUuxq93.js",
"/assets/urlFor-BU3g6tJ3.js",
"/assets/index-LUh8MTnN.js"
],
"css": [
"/assets/root-DDCa6fA8.css"
]
},
"routes/library.article.$slug": {
"id": "routes/library.article.$slug",
"parentId": "root",
"path": "library/article/:slug",
"hasAction": false,
"hasLoader": true,
"hasClientAction": false,
"hasClientLoader": false,
"hasClientMiddleware": false,
"hasErrorBoundary": false,
"module": "/assets/library.article._slug-C6Exb9Ab.js",
"imports": [
"/assets/chunk-QMGIS6GS-BqlQPwNs.js",
"/assets/index-DTAFgDfc.js",
"/assets/blogArticle-CdYFL8Rt.js",
"/assets/index-2PDpsLtL.js",
"/assets/page-DKcPV4FZ.js",
"/assets/index-DY5Afvo1.js",
"/assets/Heavybit-CW9lClJ7.js",
"/assets/index-Bm3smRr1.js",
"/assets/index-LUh8MTnN.js",
"/assets/urlFor-BU3g6tJ3.js",
"/assets/index-DCUuxq93.js"
],
"css": []
},
"routes/_index": {
"id": "routes/_index",
"parentId": "root",
"index": true,
"hasAction": false,
"hasLoader": true,
"hasClientAction": false,
"hasClientLoader": false,
"hasClientMiddleware": false,
"hasErrorBoundary": false,
"module": "/assets/_index-Dbu1V1xz.js",
"imports": [
"/assets/chunk-QMGIS6GS-BqlQPwNs.js",
"/assets/index-2PDpsLtL.js",
"/assets/Module-wcZ_2SUd.js",
"/assets/page-DKcPV4FZ.js",
"/assets/urlFor-BU3g6tJ3.js",
"/assets/Heavybit-CW9lClJ7.js",
"/assets/index-DCUuxq93.js",
"/assets/index-DY5Afvo1.js",
"/assets/PortfolioHero-fDl5iH4F.js",
"/assets/SpotlightHero-CZjr-PEe.js",
"/assets/index-LUh8MTnN.js"
],
"css": [
"/assets/Module-DnIQCapc.css"
]
},
"routes/library.$slug": {
"id": "routes/library.$slug",
"parentId": "root",
"path": "library/:slug",
"hasAction": false,
"hasLoader": true,
"hasClientAction": false,
"hasClientLoader": false,
"hasClientMiddleware": false,
"hasErrorBoundary": false,
"module": "/assets/library._slug-shJG6Dq7.js",
"imports": [
"/assets/chunk-QMGIS6GS-BqlQPwNs.js",
"/assets/index-DTAFgDfc.js",
"/assets/index-2PDpsLtL.js",
"/assets/Module-wcZ_2SUd.js",
"/assets/page-DKcPV4FZ.js",
"/assets/urlFor-BU3g6tJ3.js",
"/assets/Heavybit-CW9lClJ7.js",
"/assets/index-DCUuxq93.js",
"/assets/index-DY5Afvo1.js",
"/assets/PortfolioHero-fDl5iH4F.js",
"/assets/SpotlightHero-CZjr-PEe.js",
"/assets/index-LUh8MTnN.js"
],
"css": [
"/assets/Module-DnIQCapc.css"
]
},
"routes/$slug": {
"id": "routes/$slug",
"parentId": "root",
"path": ":slug",
"hasAction": false,
"hasLoader": true,
"hasClientAction": false,
"hasClientLoader": false,
"hasClientMiddleware": false,
"hasErrorBoundary": false,
"module": "/assets/_slug-DBEOlnLn.js",
"imports": [
"/assets/chunk-QMGIS6GS-BqlQPwNs.js",
"/assets/index-2PDpsLtL.js",
"/assets/Module-wcZ_2SUd.js",
"/assets/page-DKcPV4FZ.js",
"/assets/urlFor-BU3g6tJ3.js",
"/assets/Heavybit-CW9lClJ7.js",
"/assets/index-DCUuxq93.js",
"/assets/index-DY5Afvo1.js",
"/assets/PortfolioHero-fDl5iH4F.js",
"/assets/SpotlightHero-CZjr-PEe.js",
"/assets/index-LUh8MTnN.js"
],
"css": [
"/assets/Module-DnIQCapc.css"
]
}
},
"url": "/assets/manifest-efa36482.js",
"version": "efa36482"
};
window.__reactRouterRouteModules = {"root":route0,"routes/library.article.$slug":route1};
import("/assets/entry.client-yKU2CyKw.js");</script></body></html><!--$?--><template id="B:0"></template><!--/$--><div hidden id="S:0"><script>window.__reactRouterContext.streamController.enqueue("[{\"_1\":2,\"_5815\":-5,\"_5816\":-5},\"loaderData\",{\"_3\":4,\"_270\":271},\"root\",{\"_5\":6,\"_7\":8,\"_252\":253},\"ua\",\"Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/125.0 Safari/537.36\",\"data\",{\"_9\":10,\"_145\":146,\"_158\":159},\"footer\",{\"_11\":12,\"_13\":9,\"_14\":15,\"_16\":17,\"_23\":9,\"_24\":25,\"_26\":27,\"_34\":35,\"_107\":108},\"_createdAt\",\"2022-05-17T11:04:58Z\",\"_id\",\"_rev\",\"6sA5HXzqQ1t7bsnHMaqtNi\",\"_system\",{\"_18\":19},\"base\",{\"_20\":9,\"_21\":22},\"id\",\"rev\",\"GthKAURudFYTNgxt0CQU64\",\"_type\",\"_updatedAt\",\"2026-03-18T16:16:29Z\",\"emailSubscription\",{\"_28\":29,\"_30\":31,\"_32\":33},\"description\",\"Unlock success from technical pioneers in software infrastructure. We send the best strategies, tools, and industry updates directly to your inbox.\",\"targetURL\",\"https://share.hsforms.com/1_KYaR4lAQm-C8_9H7r4MIwqdoyu\",\"title\",\"Stay in the loop\",\"footerColumns\",[36,64,88],{\"_37\":38,\"_23\":39,\"_40\":41,\"_42\":43,\"_49\":62,\"_32\":63},\"_key\",\"30a87ef6e7c1\",\"footerColumn\",\"heading\",\"About\",\"items\",[44,52,57],{\"_37\":45,\"_23\":46,\"_47\":48,\"_49\":50,\"_32\":51},\"3dfef1283ab5\",\"footerLink\",\"label\",\"Portfolio\",\"link\",\"/portfolio\",\"Navigate to portfolio page\",{\"_37\":53,\"_23\":46,\"_47\":54,\"_49\":55,\"_32\":56},\"42666d0b3987\",\"Team\",\"/team\",\"Navigate to team page\",{\"_37\":58,\"_23\":46,\"_47\":59,\"_49\":60,\"_32\":61},\"398a9b50e214\",\"News\",\"/press\",\"Navigate to press page\",\"/about\",\"Navigate to about page\",{\"_37\":65,\"_23\":39,\"_40\":66,\"_42\":67,\"_49\":86,\"_32\":87},\"a584a6acf07b\",\"Community\",[68,72,77,82],{\"_37\":69,\"_23\":46,\"_47\":70,\"_49\":71,\"_32\":70},\"4358699f4ca4\",\"Events\",\"/events\",{\"_37\":73,\"_23\":46,\"_47\":74,\"_49\":75,\"_32\":76},\"e2d5641ef7de\",\"DevGuild\",\"https://www.heavybit.com/devguild\",\"Navigate to devguild page\",{\"_37\":78,\"_23\":46,\"_47\":79,\"_49\":80,\"_32\":81},\"feb038ce46a1\",\"Jobs\",\"/jobs\",\"Portfolio Company Jobs\",{\"_37\":83,\"_23\":46,\"_47\":84,\"_49\":85,\"_32\":84},\"e22bfa6b2330\",\"Inclusion\",\"/community/diversity-and-inclusion\",\"https://www.heavybit.com/community\",\"Navigate to community page\",{\"_37\":89,\"_23\":39,\"_40\":90,\"_42\":91,\"_49\":95,\"_32\":106},\"ea618aa34610\",\"Library\",[92,97,102],{\"_37\":93,\"_23\":46,\"_47\":94,\"_49\":95,\"_32\":96},\"c82a64edc4dd\",\"Blog\",\"https://www.heavybit.com/library/\",\"Navigate to browse on library page\",{\"_37\":98,\"_23\":46,\"_47\":99,\"_49\":100,\"_32\":101},\"e1673433290e\",\"Podcast Network\",\"https://www.heavybit.com/library/podcasts\",\"Navigate to podcasts page\",{\"_37\":103,\"_23\":46,\"_47\":104,\"_49\":105,\"_32\":104},\"96a97ceba3a2\",\"DevToolsDigest\",\"https://www.heavybit.com/devtoolsdigest\",\"Navigate to library page\",\"socialMedia\",[109,121,127,133,139],{\"_37\":110,\"_111\":112,\"_119\":120},\"ee025c96397c\",\"icon\",{\"_23\":113,\"_114\":115},\"mainImage\",\"asset\",{\"_116\":117,\"_23\":118},\"_ref\",\"image-9d618da0948299428b29e5d499af3fa5369a24d2-20x20-svg\",\"reference\",\"url\",\"https://twitter.com/heavybit\",{\"_37\":122,\"_111\":123,\"_119\":126},\"5348b028f0d41c3dac62b3dd2b5531ee\",{\"_23\":113,\"_114\":124},{\"_116\":125,\"_23\":118},\"image-01bd92eb06873bdc30b1cb5f9c78b1ef5a9559d4-512x512-png\",\"https://www.linkedin.com/company/heavybit\",{\"_37\":128,\"_111\":129,\"_119\":132},\"fae1cc0df5c6\",{\"_23\":113,\"_114\":130},{\"_116\":131,\"_23\":118},\"image-6c548d2e815d1fb3b3077c9b8be1ec251e1bc418-512x512-png\",\"https://www.youtube.com/c/heavybit\",{\"_37\":134,\"_111\":135,\"_119\":138},\"a100f6448036\",{\"_23\":113,\"_114\":136},{\"_116\":137,\"_23\":118},\"image-b38b7ee9ccd05f61ca43d5b29fe7058f822f87eb-512x512-webp\",\"https://www.instagram.com/heavybitindustries/\",{\"_37\":140,\"_111\":141,\"_119\":144},\"639729864c2d\",{\"_23\":113,\"_114\":142},{\"_116\":143,\"_23\":118},\"image-48ac9473e30341c4bee9f6634b77b6f8f1dd6b2f-849x980-webp\",\"https://www.tiktok.com/@heavybitindustries\",\"globalConfig\",{\"_147\":148,\"_11\":152,\"_13\":145,\"_14\":153,\"_16\":154,\"_23\":145,\"_24\":157},\"OGimage\",{\"_23\":149,\"_114\":150},\"image\",{\"_116\":151,\"_23\":118},\"image-32fdfae4fd13b8f9a3c5bf6b13a635068cb0ecbb-1332x529-png\",\"2022-05-17T11:15:28Z\",\"RDgKg7yO6ZYtH7Ot8nWSxU\",{\"_18\":155},{\"_20\":145,\"_21\":156},\"2ZxAYmIihJuAs2TKuKQs3J\",\"2025-08-19T13:26:15Z\",\"header\",{\"_11\":160,\"_13\":158,\"_14\":161,\"_16\":162,\"_23\":158,\"_24\":165,\"_166\":167,\"_177\":178},\"2022-05-17T11:04:33Z\",\"vtMNKfGe7wOuNuUISdyCU2\",{\"_18\":163},{\"_20\":158,\"_21\":164},\"vtMNKfGe7wOuNuUISdxyZ3\",\"2026-09-01T20:44:27Z\",\"announcementBar\",{\"_168\":169,\"_170\":171,\"_172\":173,\"_174\":175,\"_119\":176},\"buttonText\",\"RSVP\",\"show\",true,\"start\",\"2026-08-03T18:37:28.492Z\",\"text\",\"🚨 Almost sold out [Sep 24] Long-Horizon Stack conference feat Hex, Baseten, and more\",\"https://luma.com/long-horizon-stack?utm_source=heavybit_site\u0026utm_medium=banner\",\"headerNavLinks\",[179,197,207,220,248],{\"_37\":180,\"_23\":181,\"_182\":171,\"_183\":184,\"_47\":41,\"_49\":62,\"_195\":196,\"_32\":63},\"a8d641babed2\",\"navLink\",\"addDropdown\",\"dropdown\",[185,188,193],{\"_37\":186,\"_23\":187,\"_47\":54,\"_49\":55,\"_32\":56},\"964627410473\",\"dropdownLink\",{\"_37\":189,\"_23\":187,\"_47\":190,\"_49\":191,\"_32\":192},\"3d279775d73b\",\"FAQ\",\"/faq\",\"Navigate to FAQ page\",{\"_37\":194,\"_23\":187,\"_47\":59,\"_49\":60,\"_32\":61},\"84e68f9f5c33\",\"mobileOnly\",false,{\"_37\":198,\"_23\":181,\"_182\":171,\"_183\":199,\"_47\":48,\"_49\":50,\"_195\":196,\"_32\":51},\"90b68b861c06\",[200,205],{\"_37\":201,\"_23\":187,\"_47\":202,\"_49\":203,\"_32\":204},\"d481fcbbd3af\",\"Spotlights\",\"/portfolio/spotlights\",\"Navigate to spotlights page\",{\"_37\":206,\"_23\":187,\"_47\":79,\"_49\":80,\"_32\":81},\"e50fa43ad2e7\",{\"_37\":208,\"_23\":181,\"_182\":171,\"_183\":209,\"_47\":66,\"_49\":219,\"_195\":196,\"_32\":87},\"716b76469486\",[210,212,216],{\"_37\":211,\"_23\":187,\"_47\":70,\"_49\":71,\"_32\":70},\"3a177efc4649\",{\"_37\":213,\"_23\":187,\"_47\":214,\"_49\":215,\"_32\":214},\"1fb7c9c465d0\",\"Advisors\",\"/community/advisors\",{\"_37\":217,\"_23\":187,\"_47\":74,\"_49\":218,\"_32\":76},\"23077fca0e2f\",\"/devguild\",\"/community\",{\"_37\":221,\"_23\":181,\"_182\":171,\"_183\":222,\"_47\":90,\"_49\":247,\"_195\":196,\"_32\":106},\"4bec0dd665d4\",[223,228,233,236,240,244],{\"_37\":224,\"_23\":187,\"_47\":225,\"_49\":226,\"_32\":227},\"63e48d609548\",\"Articles\",\"/library?filter=articles\",\"Read educational articles in Heavybit's Library\",{\"_37\":229,\"_23\":187,\"_47\":230,\"_49\":231,\"_32\":232},\"6526c69f5260\",\"Videos\",\"/library?filter=videos\",\"Watch educational videos in Heavybit's Library\",{\"_37\":234,\"_23\":187,\"_47\":99,\"_49\":235,\"_32\":101},\"1f3fb936f3b9\",\"/library/podcasts\",{\"_37\":237,\"_23\":187,\"_47\":238,\"_49\":239,\"_32\":238},\"ac06751f0372\",\"Collections\",\"/library/collections\",{\"_37\":241,\"_23\":187,\"_47\":242,\"_49\":243,\"_32\":242},\"2ead3bd14715\",\"Reports\",\"/reports\",{\"_37\":245,\"_23\":187,\"_47\":104,\"_49\":246,\"_32\":104},\"66773779036c\",\"/devtoolsdigest\",\"/library\",{\"_37\":249,\"_23\":181,\"_182\":196,\"_47\":250,\"_49\":60,\"_195\":171,\"_32\":251},\"3702498a502c\",\"Contact\",\"Navigate to contact page\",\"ENV\",{\"_254\":255,\"_256\":257,\"_258\":259,\"_260\":261,\"_262\":263,\"_264\":265,\"_266\":267,\"_268\":255,\"_269\":255},\"SANITY_DATASET\",\"production\",\"ALGOLIA_INDEX\",\"heavybit-library\",\"ALGOLIA_APPLICATION_ID\",\"0QCKBC1FMK\",\"ALGOLIA_SITEWIDE_INDEX\",\"heavybit-sitewide\",\"ALGOLIA_JOBS_INDEX\",\"heavybit-jobs\",\"MUX_ENV_KEY\",\"dRI98bL8ze03gYDL6-Ogle2Js\",\"ALGOLIA_SEARCH_API_KEY\",\"7733fdd199ec7af7ee2923ec39c3f036\",\"MODE\",\"VITE_CONTEXT\",\"routes/library.article.$slug\",{\"_7\":272,\"_5814\":196,\"_366\":5771},{\"_273\":274},\"routeData\",{\"_11\":275,\"_13\":276,\"_14\":277,\"_23\":278,\"_24\":279,\"_280\":-5,\"_281\":282,\"_377\":378,\"_393\":1943,\"_398\":2054,\"_2042\":-5,\"_2820\":2821,\"_891\":2822,\"_2831\":2832,\"_901\":5700,\"_5707\":5708,\"_912\":5764,\"_295\":5765,\"_366\":5770,\"_920\":5772,\"_32\":5813},\"2023-10-09T22:07:52Z\",\"d285c495-1ad9-437f-9d3e-d39fd8e73869\",\"ZLU2uNGwePaNOG3gIoi3Qc\",\"blogArticle\",\"2026-08-27T23:00:18Z\",\"authors\",\"authorsWithOrg\",[283],{\"_37\":284,\"_23\":285,\"_286\":287,\"_300\":301},\"42ee258c0e02\",\"author\",\"organization\",{\"_11\":288,\"_13\":289,\"_14\":290,\"_23\":286,\"_24\":288,\"_291\":196,\"_292\":293,\"_294\":196,\"_295\":296},\"2022-07-21T16:35:12Z\",\"fb878f3a-5a88-4ee9-a3d3-927c92b46a31\",\"aeN0todGvIjkWyTywFzauG\",\"generate\",\"name\",\"Heavybit\",\"portfolioCompany\",\"seo\",{\"_23\":295,\"_28\":297,\"_298\":196,\"_299\":196,\"_32\":293},\"The leading investor in developer-first startups\",\"nofollow\",\"noindex\",\"person\",{\"_11\":302,\"_13\":303,\"_14\":304,\"_16\":305,\"_23\":300,\"_24\":308,\"_309\":310,\"_291\":171,\"_326\":327,\"_292\":332,\"_286\":333,\"_335\":171,\"_336\":337,\"_364\":365,\"_366\":367,\"_370\":371,\"_376\":171},\"2021-09-21T20:59:30Z\",\"person-1031\",\"x0IdcU5sGCNmvlYCNyWC1P\",{\"_18\":306},{\"_20\":303,\"_21\":307},\"x0IdcU5sGCNmvlYCNyW9TL\",\"2026-02-17T20:28:36Z\",\"bio\",[311],{\"_37\":312,\"_23\":313,\"_314\":315,\"_322\":323,\"_324\":325},\"7d5398a95e93\",\"block\",\"children\",[316],{\"_37\":317,\"_23\":318,\"_319\":320,\"_174\":321},\"c6b935e696790\",\"span\",\"marks\",[],\"Jesse is the co-founder of Chef and Orion Labs. He was the first chair of O’Reilly’s Velocity conference kickstarting the DevOps movement and served as Amazon’s original “Master of Disaster.” He is an avid dev community-builder and technical advisor and holds board positions with Sanity, Mobot, and others.\",\"markDefs\",[],\"style\",\"normal\",\"modules\",[328],{\"_37\":329,\"_23\":330,\"_32\":331},\"bbe24d12b0aa\",\"textBlock\",\"About Jesse Robbins\",\"Jesse Robbins\",{\"_11\":288,\"_13\":289,\"_14\":290,\"_23\":286,\"_24\":288,\"_291\":196,\"_292\":293,\"_294\":196,\"_295\":334},{\"_23\":295,\"_28\":297,\"_298\":196,\"_299\":196,\"_32\":293},\"partner\",\"photo\",{\"_23\":113,\"_338\":339,\"_114\":340,\"_342\":343,\"_353\":354},\"alt\",\"Jesse Robbins's Headshot\",{\"_116\":341,\"_23\":118},\"image-9a8f35b91b4f52c518bf668debf266feeb8e2a4a-4000x6000-jpg\",\"crop\",{\"_23\":344,\"_345\":346,\"_347\":348,\"_349\":350,\"_351\":352},\"sanity.imageCrop\",\"bottom\",0.43221712461975803,\"left\",0.2510375650388461,\"right\",0.25374915971484635,\"top\",0.13111504513040445,\"hotspot\",{\"_23\":355,\"_356\":357,\"_358\":359,\"_360\":361,\"_362\":363},\"sanity.imageHotspot\",\"height\",0.3067246795990623,\"width\",0.4095268834846873,\"x\",0.4872929063286731,\"y\",0.2969194135540817,\"position\",\"General Partner\",\"slug\",{\"_23\":366,\"_368\":369},\"current\",\"jesse-robbins\",\"socials\",[372],{\"_37\":373,\"_23\":107,\"_49\":374,\"_292\":375},\"7387c547908d\",\"https://www.linkedin.com/in/jesserobbins/\",\"linkedin\",\"teamMember\",\"browse\",[379,946,1347],{\"_11\":380,\"_13\":381,\"_14\":382,\"_16\":383,\"_23\":278,\"_24\":386,\"_281\":387,\"_393\":394,\"_398\":399,\"_891\":892,\"_901\":902,\"_912\":913,\"_295\":914,\"_366\":918,\"_920\":921,\"_943\":944,\"_32\":945},\"2024-09-05T17:26:22Z\",\"011cbf1c-cce7-4ccf-abe6-984d9aafb24d\",\"L4BYOcmxufSjuvG0B6vKrV\",{\"_18\":384},{\"_20\":381,\"_21\":385},\"sn5Y802DZciFXniDNhYgU0\",\"2025-08-19T13:26:34Z\",[388],{\"_37\":389,\"_23\":285,\"_286\":390,\"_300\":391},\"3abc96b3ac9a\",{\"_116\":289,\"_23\":118},{\"_116\":392,\"_23\":118},\"6d6d1b2d-a403-4b78-b3ea-569b1f3e9cc4\",\"collections\",[395],{\"_37\":396,\"_116\":397,\"_23\":118},\"zkrjkIbtzCb5\",\"3cd515b1-7377-4e5e-aa93-41bedfae147f\",\"content\",[400,409,417,421,441,450,458,466,474,482,490,498,515,523,531,539,543,551,559,567,575,605,613,632,656,705,713,740,785,804,808,816,824,832,856,875,883],{\"_37\":401,\"_23\":313,\"_314\":402,\"_322\":407,\"_324\":408},\"cc38b0102e2d\",[403],{\"_37\":404,\"_23\":318,\"_319\":405,\"_174\":406},\"adeb6dc1c1930\",[],\"Enterprise AI Infrastructure: Privacy, Economics, and Best First Steps\",[],\"h2\",{\"_37\":410,\"_23\":313,\"_314\":411,\"_322\":416,\"_324\":325},\"6533895d9440\",[412],{\"_37\":413,\"_23\":318,\"_319\":414,\"_174\":415},\"760b575e1f1a0\",[],\"The path to perfect AI infrastructure has yet to be paved. Enterprises must consider many important factors, like maintaining data privacy and scaling their AI deployments without overspending. They must kick off and mature their AI initiatives in a way that provides competitive advantage, then properly resource their teams.\",[],{\"_37\":418,\"_23\":149,\"_114\":419},\"756e499ad42a\",{\"_116\":420,\"_23\":118},\"image-49cd87a6d664947e706e921a10abdc1c1bf2548a-200x200-jpg\",{\"_37\":422,\"_23\":313,\"_314\":423,\"_322\":437,\"_324\":325},\"b97c5c979192\",[424,428,433],{\"_37\":425,\"_23\":318,\"_319\":426,\"_174\":427},\"2dba34a3242e0\",[],\"To scope out such challenges more clearly, we spoke with \",{\"_37\":429,\"_23\":318,\"_319\":430,\"_174\":432},\"2dba34a3242e1\",[431],\"c11c5a687abd\",\"Chaoyu Yang\",{\"_37\":434,\"_23\":318,\"_319\":435,\"_174\":436},\"2dba34a3242e2\",[],\", founder and CEO of BentoML. His startup provides a developer platform for enterprise AI teams to build and scale compound AI systems. Chaoyu previously served as a software engineer at AI data leader Databricks. Some of his key observations include:\",[438],{\"_37\":431,\"_23\":49,\"_439\":440},\"href\",\"https://www.linkedin.com/in/parano/\",{\"_37\":442,\"_23\":313,\"_314\":443,\"_322\":448,\"_324\":449},\"a40771b6363e\",[444],{\"_37\":445,\"_23\":318,\"_319\":446,\"_174\":447},\"1274ca94a6e60\",[],\"Specialized AI Systems Will Provide an Edge\",[],\"h3\",{\"_37\":451,\"_23\":313,\"_314\":452,\"_322\":457,\"_324\":325},\"b02bca4346fa\",[453],{\"_37\":454,\"_23\":318,\"_319\":455,\"_174\":456},\"a03640386dca0\",[],\"Custom AI systems optimized for a specific use case, combined with a high-quality, proprietary dataset may provide a powerful competitive advantage for enterprises that “graduate” from relying on proprietary AI models.\",[],{\"_37\":459,\"_23\":313,\"_314\":460,\"_322\":465,\"_324\":449},\"ce231cb08156\",[461],{\"_37\":462,\"_23\":318,\"_319\":463,\"_174\":464},\"834f691c74b10\",[],\"Data Privacy in AI Will Be Crucial, Especially for Highly Regulated Spaces\",[],{\"_37\":467,\"_23\":313,\"_314\":468,\"_322\":473,\"_324\":325},\"2b1f15de9b9f\",[469],{\"_37\":470,\"_23\":318,\"_319\":471,\"_174\":472},\"f691e4b8d2b50\",[],\"As enterprises scale their ML workloads, data privacy will only become more important, particularly as each company’s store of proprietary data grows in size and relevance.\",[],{\"_37\":475,\"_23\":313,\"_314\":476,\"_322\":481,\"_324\":449},\"b30f62567a7b\",[477],{\"_37\":478,\"_23\":318,\"_319\":479,\"_174\":480},\"04abface7e720\",[],\"Future Economic Shifts May Lead to an Inflection Point\",[],{\"_37\":483,\"_23\":313,\"_314\":484,\"_322\":489,\"_324\":325},\"e76511d604f7\",[485],{\"_37\":486,\"_23\":318,\"_319\":487,\"_174\":488},\"6f1e4013aeeb0\",[],\"While there’s a case to be made for every enterprise to run and own their own AI/ML operations internally, it’s arguably not a practical goal at the moment due to a variety of factors (including operational gaps between data science and operations teams, and the sheer economics of trying to own your own inference estate). In the future, more tools and cheaper, commoditized compute resources may eventually tip the scales in a way that makes owning their own AI operations more feasible for enterprises.\",[],{\"_37\":491,\"_23\":313,\"_314\":492,\"_322\":497,\"_324\":408},\"698a8c5fd001\",[493],{\"_37\":494,\"_23\":318,\"_319\":495,\"_174\":496},\"fb949f9f43660\",[],\"Maturity and How to Scale\",[],{\"_37\":499,\"_23\":313,\"_314\":500,\"_322\":514,\"_324\":325},\"237788f66736\",[501,505,510],{\"_37\":502,\"_23\":318,\"_319\":503,\"_174\":504},\"ed23ba01eca60\",[],\"While the current paradigm for inference hosting boils down to some combination of on-device, hosted, or data center, deployment and inference platform each have their own nuances–though there’s an emerging platform ecosystem for inference hosting as well. “There are a number of players providing an inference \",{\"_37\":506,\"_23\":318,\"_319\":507,\"_174\":509},\"ed23ba01eca61\",[508],\"em\",\"platform\",{\"_37\":511,\"_23\":318,\"_319\":512,\"_174\":513},\"ed23ba01eca62\",[],\". There’s my own team at BentoML and other providers that let developers deploy any open-source model, or their own proprietary custom model, and run inference at scale. Compared to AI API providers, a big differentiator is that we offer dedicated ‘bring your own cloud’-style private deployment, typically in customers’ own secured environment–which is especially appropriate for teams in highly regulated industries.”\",[],{\"_37\":516,\"_23\":313,\"_314\":517,\"_322\":522,\"_324\":325},\"347d4b74d16a\",[518],{\"_37\":519,\"_23\":318,\"_319\":520,\"_174\":521},\"46f6a21068bf0\",[],\"Chaoyu suggests a maturity curve that begins with hosted options due to ease of use through providers like OpenAI and Anthropic. “For now, people mostly get started with an API endpoint provider. There's an argument that self-hosting AI models can provide cost benefits as you scale up, but we found that to be a weak argument. Even teams using GPT-4 at scale, especially for enterprise users, don't think of it as ‘expensive.’ Especially when compared to the total cost of ownership of building and maintaining your own mission-critical AI systems.”\",[],{\"_37\":524,\"_23\":313,\"_314\":525,\"_322\":530,\"_324\":325},\"99ebcbb3b598\",[526],{\"_37\":527,\"_23\":318,\"_319\":528,\"_174\":529},\"9ffc91be95d10\",[],\"The CEO suggests that early adopters are starting to look for alternatives. “As AI increasingly powers your business-critical applications, specialized AI systems are going to become more strategically important for every enterprise to compete. Being able to build specialized AI systems, optimize for your specific business use case and leverage proprietary data and knowledge, are going to become how you win in the future.”\",[],{\"_37\":532,\"_23\":313,\"_314\":533,\"_322\":538,\"_324\":325},\"1f8a57c57461\",[534],{\"_37\":535,\"_23\":318,\"_319\":536,\"_174\":537},\"d034792b46cb0\",[],\"“Another important topic is developer efficiency. For a lot of use cases, you will need to quickly iterate on either the application code, the model, the inference strategy, or infrastructure decisions, to tweak system design and make performance improvements. ‘Performance’ could include factors like latency or accuracy for a specific scenario, cost efficiency, or security metrics. Today, people are still racing to get their product to market, to start evaluating ROI, to start evaluating how AI is contributing to the business. But over time, I think the specialized-AI approach will win in high impact, business critical AI applications.”\",[],{\"_37\":540,\"_23\":149,\"_114\":541},\"82c00b47bd13\",{\"_116\":542,\"_23\":118},\"image-172e3e37deab2e7c88817ff86bb8c29441bdadf1-2500x877-png\",{\"_37\":544,\"_23\":313,\"_314\":545,\"_322\":550,\"_324\":325},\"f67e98ebad78\",[546],{\"_37\":547,\"_23\":318,\"_319\":548,\"_174\":549},\"cfc2f73468430\",[508],\"Data centers require potentially massive total cost of ownership, from real estate to hardware to ongoing maintenance. Image courtesy MIT Technology Review\",[],{\"_37\":552,\"_23\":313,\"_314\":553,\"_322\":558,\"_324\":408},\"499e18b50dc3\",[554],{\"_37\":555,\"_23\":318,\"_319\":556,\"_174\":557},\"02dbe2a59117\",[],\"From First Principles to Enterprise Adoption\",[],{\"_37\":560,\"_23\":313,\"_314\":561,\"_322\":566,\"_324\":325},\"b32000c6ed5f\",[562],{\"_37\":563,\"_23\":318,\"_319\":564,\"_174\":565},\"f2e7fbdac20b0\",[],\"Even with this potential future in mind, Chaoyu would still likely advise newcomers to consider starting with a third-party AI API provider. “That's just the fastest way to explore what AI could do for you. If you're working with sensitive data, try to curate synthetic data that's less sensitive for prototyping and evaluation.”\",[],{\"_37\":568,\"_23\":313,\"_314\":569,\"_322\":574,\"_324\":325},\"93c30f32d7c9\",[570],{\"_37\":571,\"_23\":318,\"_319\":572,\"_174\":573},\"6beeb6b4cdab0\",[],\"“As enterprises look to build specialized AI systems however, they should carefully consider TCO, which can be prohibitively high. To mature over time, enterprises require platforms to help scale their models and empower application building. But I do think a couple trends are pointing towards more adoption of specialized AI systems. The first trend is on-demand GPUs becoming more accessible, at cheaper prices. I think it will consolidate to a few cloud vendors that will make on-demand GPU access fairly easy in the next 12 months. A second trend is open-source models. You have tons of really good options that open up opportunities for advanced customizations, and they're only getting better from here.”\",[],{\"_37\":576,\"_23\":313,\"_314\":577,\"_322\":600,\"_324\":325},\"dc2bf1656874\",[578,582,587,591,596],{\"_37\":579,\"_23\":318,\"_319\":580,\"_174\":581},\"13d379b9015c0\",[],\"“And the last trend is the growing shift to \",{\"_37\":583,\"_23\":318,\"_319\":584,\"_174\":586},\"13d379b9015c1\",[585,508],\"6ab979b225d1\",\"compound AI systems\",{\"_37\":588,\"_23\":318,\"_319\":589,\"_174\":590},\"13d379b9015c2\",[],\"–systems that utilize multiple interacting components, models, and other tools to accomplish their goals. I think the future won’t be a single model that does everything. More AI applications are compound systems composed of multiple models, multiple pipelines, and multiple components. \",{\"_37\":592,\"_23\":318,\"_319\":593,\"_174\":595},\"13d379b9015c3\",[594],\"c022ba38ce5e\",\"RAG\",{\"_37\":597,\"_23\":318,\"_319\":598,\"_174\":599},\"13d379b9015c4\",[],\", Voice Chat LLM, Function-calling Agents are some of the popular examples. And this is a trend that will result in more AI applications being built with a combination of foundation models and specialized models, and need access to sensitive data or proprietary software systems. We’ll return to this topic shortly.”\",[601,603],{\"_37\":585,\"_23\":49,\"_439\":602},\"https://bair.berkeley.edu/blog/2024/02/18/compound-ai-systems/\",{\"_37\":594,\"_23\":49,\"_439\":604},\"https://en.wikipedia.org/wiki/Rag\",{\"_37\":606,\"_23\":313,\"_314\":607,\"_322\":612,\"_324\":408},\"fb8a3f65383e\",[608],{\"_37\":609,\"_23\":318,\"_319\":610,\"_174\":611},\"ba07b3eb7e830\",[],\"Resource Allocation and Efficiency in AI Deployment\",[],{\"_37\":614,\"_23\":313,\"_314\":615,\"_322\":629,\"_324\":325},\"6a6ba165e2da\",[616,620,625],{\"_37\":617,\"_23\":318,\"_319\":618,\"_174\":619},\"2d93ae6a106d0\",[],\"The founder notes that resource utilization is a surprising challenge at many ends of the spectrum for enterprises. “You noted in a \",{\"_37\":621,\"_23\":318,\"_319\":622,\"_174\":624},\"2d93ae6a106d1\",[623],\"dc02b357e51b\",\"blog post about misaligned AI incentives\",{\"_37\":626,\"_23\":318,\"_319\":627,\"_174\":628},\"2d93ae6a106d2\",[],\" that only 10% of AI projects go into production. Obviously, things should be aligned in terms of business goals, and there are some similarities [to the pre-DevOps days] where engineers and ML scientists have a very different path and different day-to-day outputs. However, I believe there should be much better tooling designed to close some of those gaps in the future.”\",[630],{\"_37\":623,\"_23\":49,\"_439\":631},\"https://www.heavybit.com/library/article/machine-learning-engineering-ai-incentives\",{\"_37\":633,\"_23\":313,\"_314\":634,\"_322\":655,\"_324\":325},\"69ec0083e663\",[635,639,643,647,651],{\"_37\":636,\"_23\":318,\"_319\":637,\"_174\":638},\"ceb3a1d21024\",[],\"“Today, a lot of AI projects \",{\"_37\":640,\"_23\":318,\"_319\":641,\"_174\":642},\"63a94cd274391\",[508],\"say\",{\"_37\":644,\"_23\":318,\"_319\":645,\"_174\":646},\"63a94cd274392\",[],\" they are in production and paying a fortune for GPU resources. But they don’t know the exact resource \",{\"_37\":648,\"_23\":318,\"_319\":649,\"_174\":650},\"63a94cd274393\",[508],\"utilization\",{\"_37\":652,\"_23\":318,\"_319\":653,\"_174\":654},\"63a94cd274394\",[],\" rate, oftentimes struggling with GPU utilization at under 10%.” The CEO suggests that resource overprovision and its upfront cost, is another argument in favor of newer AI projects beginning with endpoint providers.\",[],{\"_37\":657,\"_23\":313,\"_314\":658,\"_322\":698,\"_324\":325},\"de31e7e15412\",[659,663,667,671,676,680,685,689,694],{\"_37\":660,\"_23\":318,\"_319\":661,\"_174\":662},\"4f7603186c020\",[],\"“For example, within an enterprise, I may reserve a limited amount of GPU quota, dynamically shared with dozens, or hundreds, of models, making sure inactive models are being scaled down to zero and high priority models are right-sized to ensure service quality. In my mind, \",{\"_37\":664,\"_23\":318,\"_319\":665,\"_174\":666},\"4f7603186c021\",[508],\"that’s\",{\"_37\":668,\"_23\":318,\"_319\":669,\"_174\":670},\"4f7603186c022\",[],\" the hard infrastructure problem we aim to solve.” Chaoyu offers the example of serving open-source LLMs, for which teams might use \",{\"_37\":672,\"_23\":318,\"_319\":673,\"_174\":675},\"4f7603186c023\",[674],\"b5318d663b9b\",\"vLLM\",{\"_37\":677,\"_23\":318,\"_319\":678,\"_174\":679},\"4f7603186c024\",[],\", \",{\"_37\":681,\"_23\":318,\"_319\":682,\"_174\":684},\"4f7603186c025\",[683],\"83ff407adfba\",\"text-generated inference\",{\"_37\":686,\"_23\":318,\"_319\":687,\"_174\":688},\"4f7603186c026\",[],\", or \",{\"_37\":690,\"_23\":318,\"_319\":691,\"_174\":693},\"4f7603186c027\",[692],\"5161b1c794ec\",\"TensorRT-LLM\",{\"_37\":695,\"_23\":318,\"_319\":696,\"_174\":697},\"4f7603186c028\",[],\"–which all promises inference performance on a single model replica. However, the surrounding infrastructure for fast scaling, cold-start, concurrency control, observability, and common LLM deployment patterns–such as LLM Guardrails, Multi-LLM gateway–can still be quite difficult to build and optimize.\",[699,701,703],{\"_37\":674,\"_23\":49,\"_439\":700},\"https://github.com/vllm-project/vllm\",{\"_37\":683,\"_23\":49,\"_439\":702},\"https://huggingface.co/docs/text-generation-inference/en/index\",{\"_37\":692,\"_23\":49,\"_439\":704},\"https://github.com/NVIDIA/TensorRT-LLM\",{\"_37\":706,\"_23\":313,\"_314\":707,\"_322\":712,\"_324\":325},\"c46e559ff53a\",[708],{\"_37\":709,\"_23\":318,\"_319\":710,\"_174\":711},\"4cb043a9bc6a0\",[],\"Traditional Cloud-native infrastructure doesn’t really work for GenAI –container images and models files are huge, for instance, leading to delays of 30 minutes or more when scaling from one instance to two. “You can see how a lot of the traditional DevOps assumptions about the workload being a small-container, single-process type of thing–how GenAI just completely breaks that. We need a new type of cloud infrastructure, optimizing every step in the AI stack, to solve the resource efficiency challenge.”\",[],{\"_37\":714,\"_23\":313,\"_314\":715,\"_322\":737,\"_324\":325},\"0b8106fa896f\",[716,720,725,729,733],{\"_37\":717,\"_23\":318,\"_319\":718,\"_174\":719},\"de25d4c4cf0c0\",[],\"Engineering teams that are new to running AI in production may make assumptions about how metrics work that cause them to unwittingly run up massive usage bills, or hit performance snags as jobs fail or slow to a crawl due to unexpected variations in data payloads. “People who work in DevOps may focus on straightforward CPU/GPU Kubernetes metrics but there are limitations such as Python’s \",{\"_37\":721,\"_23\":318,\"_319\":722,\"_174\":724},\"de25d4c4cf0c1\",[723,508],\"e4b3956253de\",\"global interpreter lock\",{\"_37\":726,\"_23\":318,\"_319\":727,\"_174\":728},\"de25d4c4cf0c2\",[],\", which limits parallel execution and makes CPU utilization less visible. There are also nuances in terms of how vendors like Nvidia represent utilization. And another nuance that we find newer teams can overlook is how resource-based metrics are always retrospective. You’re looking at how usage has \",{\"_37\":730,\"_23\":318,\"_319\":731,\"_174\":732},\"de25d4c4cf0c3\",[508],\"been\",{\"_37\":734,\"_23\":318,\"_319\":735,\"_174\":736},\"de25d4c4cf0c4\",[],\" consumed over a certain time period–not necessarily something that gets conveyed in a single snapshot in time.”\",[738],{\"_37\":723,\"_23\":49,\"_439\":739},\"https://wiki.python.org/moin/GlobalInterpreterLock\",{\"_37\":741,\"_23\":313,\"_314\":742,\"_322\":778,\"_324\":325},\"ed9d0ad6e66e\",[743,747,752,756,760,765,769,774],{\"_37\":744,\"_23\":318,\"_319\":745,\"_174\":746},\"7d60d07936920\",[],\"For example, Chaoyu recommends considering \",{\"_37\":748,\"_23\":318,\"_319\":749,\"_174\":751},\"7d60d07936921\",[750,508],\"2159899982b6\",\"concurrency-based scaling\",{\"_37\":753,\"_23\":318,\"_319\":754,\"_174\":755},\"7d60d07936922\",[750],\",\",{\"_37\":757,\"_23\":318,\"_319\":758,\"_174\":759},\"7d60d07936923\",[],\" which can offer more-granular information that will help teams deduce how many GPUs needed to support their various workloads. “Concurrency scaling works pretty well with inference where batching could be happening. For example, you can do \",{\"_37\":761,\"_23\":318,\"_319\":762,\"_174\":764},\"7d60d07936924\",[763],\"042219c7dc02\",\"adaptive batching\",{\"_37\":766,\"_23\":318,\"_319\":767,\"_174\":768},\"7d60d07936925\",[],\" or \",{\"_37\":770,\"_23\":318,\"_319\":771,\"_174\":773},\"7d60d07936926\",[772],\"26342b0eeacf\",\"continuous batching\",{\"_37\":775,\"_23\":318,\"_319\":776,\"_174\":777},\"7d60d07936927\",[],\" that groups multiple incoming requests together.”\",[779,781,783],{\"_37\":750,\"_23\":49,\"_439\":780},\"https://docs.aws.amazon.com/lambda/latest/dg/lambda-concurrency.html#understanding-concurrency\",{\"_37\":763,\"_23\":49,\"_439\":782},\"https://docs.bentoml.com/en/latest/guides/adaptive-batching.html\",{\"_37\":772,\"_23\":49,\"_439\":784},\"https://x.com/zhuohan123/status/1671234707206590464\",{\"_37\":786,\"_23\":313,\"_314\":787,\"_322\":801,\"_324\":325},\"96d1d859aae5\",[788,792,797],{\"_37\":789,\"_23\":318,\"_319\":790,\"_174\":791},\"d13256a226800\",[],\"While it can be tempting for longtime infrastructure engineers to toss their AI workloads into K8S and expect everything to work the way cloud workloads typically do, the founder cautions that AI workloads are drastically different. “Without a proper AI infrastructure, you may end up with a larger bill from AWS than you thought because your system gets overprovisioned, or your deployments don't scale up fast enough so requests fail or respond slowly. Slow performance can be a serious issue for some use cases that need low latency, such as an AI phone calling agent. As we covered in one of \",{\"_37\":793,\"_23\":318,\"_319\":794,\"_174\":796},\"d13256a226801\",[795],\"cca132555f7d\",\"our blogs\",{\"_37\":798,\"_23\":318,\"_319\":799,\"_174\":800},\"d13256a226802\",[],\", the typical cloud-native stack is not built for AI, and can lead to suboptimal performance or utilization.\",[802],{\"_37\":795,\"_23\":49,\"_439\":803},\"https://www.bentoml.com/blog/scaling-ai-model-deployment\",{\"_37\":805,\"_23\":149,\"_114\":806},\"223808d8eece\",{\"_116\":807,\"_23\":118},\"image-4a07919ff37dfdb620a8c069295dd8f9651d50f4-958x276-png\",{\"_37\":809,\"_23\":313,\"_314\":810,\"_322\":815,\"_324\":325},\"39421985ba3f\",[811],{\"_37\":812,\"_23\":318,\"_319\":813,\"_174\":814},\"5dc9792ffdb5\",[508],\"Enterprises like Google run AI programs from massive data centers built to provide compute resources at scale. Image courtesy Forbes.\",[],{\"_37\":817,\"_23\":313,\"_314\":818,\"_322\":823,\"_324\":408},\"d7de8afed0cd\",[819],{\"_37\":820,\"_23\":318,\"_319\":821,\"_174\":822},\"91abaacee3c80\",[],\"In the Future, Enterprises Will Increasingly Focus on Compound AI\",[],{\"_37\":825,\"_23\":313,\"_314\":826,\"_322\":831,\"_324\":325},\"4768c94198e1\",[827],{\"_37\":828,\"_23\":318,\"_319\":829,\"_174\":830},\"220112ee71440\",[],\"“As I mentioned earlier, I believe that the best AI products are built with the compound AI approach, and it will become much more common in enterprise AI in the future.” Chaoyu offers. “The complexity in such systems is increasing, which makes it more important for teams to stay agile and ship faster as they unearth more use cases.”\",[],{\"_37\":833,\"_23\":313,\"_314\":834,\"_322\":855,\"_324\":325},\"ed912f3899da\",[835,839,843,847,851],{\"_37\":836,\"_23\":318,\"_319\":837,\"_174\":838},\"cb69c25d423c0\",[],\"The founder confides that despite the enormous amount of capital expenditure that larger orgs have invested into AI so far, his company’s enterprise customers aren’t fretting about the costs so much as they are concerned about developer efficiency and data privacy. “From our perspective, the main issues are \",{\"_37\":840,\"_23\":318,\"_319\":841,\"_174\":842},\"cb69c25d423c1\",[508],\"control\",{\"_37\":844,\"_23\":318,\"_319\":845,\"_174\":846},\"cb69c25d423c2\",[],\" and \",{\"_37\":848,\"_23\":318,\"_319\":849,\"_174\":850},\"cb69c25d423c3\",[508],\"customization\",{\"_37\":852,\"_23\":318,\"_319\":853,\"_174\":854},\"cb69c25d423c4\",[],\". In this case, ‘control,’ includes things like data privacy and security, avoiding vendor lock-in, and having predictable behavior. Whereas ‘customization’ would refer to the flexibility and ease of use in building out advanced compound AI systems with custom requirements.”\",[],{\"_37\":857,\"_23\":313,\"_314\":858,\"_322\":872,\"_324\":325},\"24e730cfbe99\",[859,863,868],{\"_37\":860,\"_23\":318,\"_319\":861,\"_174\":862},\"17a4997e4ffa0\",[],\"“I’ll give you an example–we have a customer building a voice agent application with multiple \",{\"_37\":864,\"_23\":318,\"_319\":865,\"_174\":867},\"17a4997e4ffa1\",[866],\"a207cb1b75a6\",\"open-source\",{\"_37\":869,\"_23\":318,\"_319\":870,\"_174\":871},\"17a4997e4ffa2\",[],\" models, including components such as speech recognition, LLM, function calling, and text-to-speech. Their initial prototype can take over a minute to respond to a user’s question, which is not acceptable in real-time voice assistant use cases.”\",[873],{\"_37\":866,\"_23\":49,\"_439\":874},\"https://www.heavybit.com/library/collections/open-source-for-startups-guide\",{\"_37\":876,\"_23\":313,\"_314\":877,\"_322\":882,\"_324\":325},\"728bd386f976\",[878],{\"_37\":879,\"_23\":318,\"_319\":880,\"_174\":881},\"49f96cba3de50\",[],\"“With our platform, they were able to quickly fine-tune their inference setup for faster time-to-first-token latency, parallelize multiple inference calls, and replace a large number of slow LLM calls with faster, domain specific models. We were able to stand up a solution that got them the ultra-low-latency deployment they needed, improving end-to-end latency to less than 1 second. In these cases, optimizations can be highly specific to the use case. That’s what we’re seeing enterprises ask for.”\",[],{\"_37\":884,\"_23\":313,\"_314\":885,\"_322\":890,\"_324\":325},\"0cc9e2a1d23a\",[886],{\"_37\":887,\"_23\":318,\"_319\":888,\"_174\":889},\"702f08bf71f20\",[],\"“I strongly recommend considering specialized AI with the compound AI systems approach, as this offers the flexibility for quickly improving performance for your specific use case.” says the CEO. “When you’re evaluating the ROI of your AI initiatives–prototypes don’t often focus on details like the latency vs. throughput tradeoffs, but when you want to productionize that into a reliable and scalable product, that becomes a lot more impactful in customer experience and cost saving.”\",[],\"excerpt\",[893],{\"_37\":894,\"_23\":313,\"_314\":895,\"_322\":900,\"_324\":325},\"2de1ba2cf88b\",[896],{\"_37\":897,\"_23\":318,\"_319\":898,\"_174\":899},\"c30ab14e37a70\",[],\"Planning enterprise infrastructure successfully means considering privacy, resourcing, and what’s to come. BentoML founder Chaoyu Yang explains.\",[],\"moreFromLibrary\",[903,906,909],{\"_37\":904,\"_116\":905,\"_23\":118},\"L88lqk4cKlKS\",\"efc727d0-1f15-4252-8a26-aa743d7761f2\",{\"_37\":907,\"_116\":908,\"_23\":118},\"ID60gLuzl0zB\",\"4e6b9cf7-5c78-4ec1-811b-8ed75be1d84a\",{\"_37\":910,\"_116\":911,\"_23\":118},\"OPfqvBrFGGoJ\",\"b229cbc7-04ac-4d25-a28e-9a02ea7fae6c\",\"publishDate\",\"2024-09-05T17:26:00.000Z\",{\"_147\":915,\"_23\":295,\"_28\":899,\"_298\":196,\"_299\":196},{\"_23\":149,\"_114\":916},{\"_116\":917,\"_23\":118},\"image-582647e4c6d0f71290062add3e8da0108936da4b-1200x630-png\",{\"_23\":366,\"_368\":919},\"enterprise-ai-infrastructure-privacy-maturity-resources\",\"tags\",[922,925,928,931,934,937,940],{\"_37\":923,\"_116\":924,\"_23\":118},\"a8b9a5230a96\",\"1cf431af-c37e-4932-872e-b2bce7cbd4ac\",{\"_37\":926,\"_116\":927,\"_23\":118},\"b6063f8520a3\",\"35d87736-5801-4207-b96c-abf2f3cabf25\",{\"_37\":929,\"_116\":930,\"_23\":118},\"0667f4011bc4\",\"sGpx8ZAtjp3y9Udp02EBLo\",{\"_37\":932,\"_116\":933,\"_23\":118},\"f834485b4887\",\"017fa9be-32e2-46ee-bddf-f13275fbdf00\",{\"_37\":935,\"_116\":936,\"_23\":118},\"5e5a70526f9f\",\"ce8ede15-ce64-431d-9cbd-d81338d1be6a\",{\"_37\":938,\"_116\":939,\"_23\":118},\"7ad250a19c7f\",\"kFpYkfzWCiYI8xC60oUpqE\",{\"_37\":941,\"_116\":942,\"_23\":118},\"97a6372a2b2a\",\"rec3De7lEeQV6Zy89\",\"templateType\",\"default\",\"Enterprise AI Infrastructure: Privacy, Maturity, Resources\",{\"_11\":947,\"_13\":948,\"_14\":949,\"_23\":278,\"_24\":950,\"_281\":951,\"_393\":957,\"_398\":961,\"_891\":1320,\"_901\":1329,\"_912\":1339,\"_295\":1340,\"_366\":1345,\"_943\":944,\"_32\":1344},\"2026-05-11T17:55:45Z\",\"0472b492-68d5-4519-a3eb-bd8db4cdd1b7\",\"uDD4M8vz3H39N7Kfu0sAna\",\"2026-06-10T01:04:53Z\",[952],{\"_37\":953,\"_23\":285,\"_286\":954,\"_300\":955},\"83597e3658d0\",{\"_116\":289,\"_23\":118},{\"_116\":956,\"_23\":118},\"person-28\",[958],{\"_37\":959,\"_116\":960,\"_23\":118},\"0hljwQD6rP28\",\"5c63b811-83c5-452e-8d3d-af8c1df55298\",[962,970,1005,1035,1043,1051,1055,1071,1079,1087,1106,1125,1133,1141,1149,1157,1165,1173,1181,1189,1197,1205,1213,1221,1229,1237,1245,1253,1261,1288,1296,1304,1312],{\"_37\":963,\"_23\":313,\"_314\":964,\"_322\":969,\"_324\":408},\"cfaab6768243\",[965],{\"_37\":966,\"_23\":318,\"_319\":967,\"_174\":968},\"8760054a996b\",[],\"How We Went from LLMs to Agents to Skills\",[],{\"_37\":971,\"_23\":313,\"_314\":972,\"_322\":1002,\"_324\":325},\"2ad3deb4acb5\",[973,977,981,985,990,994,998],{\"_37\":974,\"_23\":318,\"_319\":975,\"_174\":976},\"92b89fe4dab0\",[],\"In late 2025, Anthropic declared \",{\"_37\":978,\"_23\":318,\"_319\":979,\"_174\":980},\"86eac63ebab3\",[508],\"skills\",{\"_37\":982,\"_23\":318,\"_319\":983,\"_174\":984},\"a1c17360f8af\",[],\", modular, reusable, file-based instructions that encapsulate domain expertise, to be an \",{\"_37\":986,\"_23\":318,\"_319\":987,\"_174\":989},\"af336c0c1b6a\",[988],\"57b77040cdf1\",\"open standard\",{\"_37\":991,\"_23\":318,\"_319\":992,\"_174\":993},\"85925c92fead\",[],\". Since then, hundreds of thousands of users have reportedly installed a variety of \",{\"_37\":995,\"_23\":318,\"_319\":996,\"_174\":997},\"f1f35745072a\",[508],\"skill.md\",{\"_37\":999,\"_23\":318,\"_319\":1000,\"_174\":1001},\"696333c57e56\",[],\" files to improve their agents’ abilities on a variety of tasks, including coding with specialized languages or performing non-coding tasks like spinning up documentation or creating datasets.\",[1003],{\"_37\":988,\"_23\":49,\"_439\":1004},\"https://agentskills.io/home\",{\"_37\":1006,\"_23\":313,\"_314\":1007,\"_322\":1030,\"_324\":325},\"a85990e60299\",[1008,1012,1017,1021,1026],{\"_37\":1009,\"_23\":318,\"_319\":1010,\"_174\":1011},\"d65aec64a472\",[],\"While performing his usual daily duties with Claude Code, Turkish developer \",{\"_37\":1013,\"_23\":318,\"_319\":1014,\"_174\":1016},\"6b1fbdb6a300\",[1015],\"e7c7ecb6c450\",\"Yusuf Karaaslan\",{\"_37\":1018,\"_23\":318,\"_319\":1019,\"_174\":1020},\"89cfdbd59a97\",[],\" began experimenting with skills, and eventually ending up designing the \",{\"_37\":1022,\"_23\":318,\"_319\":1023,\"_174\":1025},\"6448583554f5\",[1024],\"e50d55cba43f\",\"Skill Seekers\",{\"_37\":1027,\"_23\":318,\"_319\":1028,\"_174\":1029},\"ef35bae225c6\",[],\" project, which processes and translates a variety of data sources (including GitHub repos, PDFs, videos, and others) into agentic skills. Below, he explains what led him to create the project and how thousands of users worldwide are finding value from agentic skills.\",[1031,1033],{\"_37\":1015,\"_23\":49,\"_439\":1032},\"https://www.linkedin.com/in/yusuf-karaaslan-156125145/\",{\"_37\":1024,\"_23\":49,\"_439\":1034},\"https://github.com/yusufkaraaslan/Skill_Seekers\",{\"_37\":1036,\"_23\":313,\"_314\":1037,\"_322\":1042,\"_324\":408},\"cad621bad29d\",[1038],{\"_37\":1039,\"_23\":318,\"_319\":1040,\"_174\":1041},\"bda5cad6a37c\",[],\"From Context Limits to Skill Libraries\",[],{\"_37\":1044,\"_23\":313,\"_314\":1045,\"_322\":1050,\"_324\":325},\"287cb7f1f994\",[1046],{\"_37\":1047,\"_23\":318,\"_319\":1048,\"_174\":1049},\"c8fc0b9f3f61\",[],\"Karaaslan, a game developer by trade, notes that his initial foray into working with skills came from trying to work with the back-end of the popular online game storefront Steam. He started with Claude's official skill-creator skill, feeding it every relevant link from Steam's documentation one by one and asking it to produce a complete skill for the inventory system — the goal being a direct answer instead of a guess. It didn't work.\",[],{\"_37\":1052,\"_23\":149,\"_114\":1053},\"197dcbe20fcb\",{\"_116\":1054,\"_23\":118},\"image-d7f4e82d7a2e0b63380beac2c7e0b663f4b503de-1749x803-jpg\",{\"_37\":1056,\"_23\":313,\"_314\":1057,\"_322\":1070,\"_324\":325},\"ff8899b3edf1\",[1058,1062,1066],{\"_37\":1059,\"_23\":318,\"_319\":1060,\"_174\":1061},\"3f01e43f5ee2\",[],\"\\\"No matter what I did, Claude would skip most of the input or cut it down far too aggressively, and it kept guessing at what I was asking for. My read was that because I'd asked it to handle both the heavy data gathering \",{\"_37\":1063,\"_23\":318,\"_319\":1064,\"_174\":1065},\"5104564a4efd\",[508],\"and\",{\"_37\":1067,\"_23\":318,\"_319\":1068,\"_174\":1069},\"4b2e01600383\",[],\" the processing in one shot, it was optimizing for cost and cutting corners on the part that mattered most. Even the cut-down output was a little more confident than nothing, which was the clue. So I sat down and started building a separate system to do the heavy data lifting, the part that doesn't actually need AI in the loop. That was the first prototype of Skill Seekers.\\\"\",[],{\"_37\":1072,\"_23\":313,\"_314\":1073,\"_322\":1078,\"_324\":325},\"db23182db83f\",[1074],{\"_37\":1075,\"_23\":318,\"_319\":1076,\"_174\":1077},\"590dd25f6d61\",[],\"The contrast once he piped that prototype's output back into Claude was hard to miss. \\\"Before, the answers were always hedged. 'Maybe try this, maybe try that.' After I added the Skill Seekers output to the context, the tone changed completely: 'You can't do it the way you want with Steam inventory, you need an external backend.' Much more confident, much more direct, and actually correct. Even when I challenged it on details, it stayed true to Steam's actual behavior. The funny part is that all this happened the same day Anthropic announced skills. The first version of Skill Seekers shipped that same night.”\",[],{\"_37\":1080,\"_23\":313,\"_314\":1081,\"_322\":1086,\"_324\":325},\"ae1630ad1dd9\",[1082],{\"_37\":1083,\"_23\":318,\"_319\":1084,\"_174\":1085},\"2177d179b453\",[],\"From there, Karaaslan kept iterating on the prototype, prompting Claude Code to extend it with web scraping and working around internal context limits until each new data source slotted in cleanly. \\\"Until a certain point, my focus was on feeding in as many data sources as possible to build a richer, more complex skill: The need to download a dynamic website, or the need to scrape a codebase. Once I felt we'd covered almost all the major data sources and some of the more niche ones, I started backtracking to improve the system architecture. At that point I hit Claude's internal limits again: This time around analyzing the whole codebase. So I fell back on my usual trick: I had Claude generate full class diagrams, package diagrams, and flow diagrams for the key systems. With those UMLs in hand, I could pinpoint exact pain points, architectural problems, and more. From there, I designed the new architecture end-to-end and replaced the live version module by module.”\",[],{\"_37\":1088,\"_23\":313,\"_314\":1089,\"_322\":1103,\"_324\":325},\"d5754bf4c7a2\",[1090,1094,1099],{\"_37\":1091,\"_23\":318,\"_319\":1092,\"_174\":1093},\"82fc7ba2330e\",[],\"The developer’s iterative approach led him to map all connections and flows within the project, and eventually unify everything with a single interface, ensuring key functions were connected and removing duplicate branches. He explains that he relies on three primary skills himself for design, implementation, and PR reviews. \\\"I actually developed these skills side by side with Skill Seekers itself, based on my own observations along the way. And I just open-sourced my \",{\"_37\":1095,\"_23\":318,\"_319\":1096,\"_174\":1098},\"26606e121eb2\",[1097],\"015765e5dc10\",\"AI workflows\",{\"_37\":1100,\"_23\":318,\"_319\":1101,\"_174\":1102},\"372f19f62b1d\",[],\" too.”\",[1104],{\"_37\":1097,\"_23\":49,\"_439\":1105},\"https://github.com/yusufkaraaslan/ai-flow-anything\",{\"_37\":1107,\"_23\":313,\"_314\":1108,\"_322\":1122,\"_324\":325},\"1194e4a8af2c\",[1109,1113,1118],{\"_37\":1110,\"_23\":318,\"_319\":1111,\"_174\":1112},\"d8e709e56465\",[],\"“Using my design skill, I start by using a card that simply explains \",{\"_37\":1114,\"_23\":318,\"_319\":1115,\"_174\":1117},\"1fbc04ff1672\",[1116],\"4860203c9541\",\"test-driven development\",{\"_37\":1119,\"_23\":318,\"_319\":1120,\"_174\":1121},\"becfe24d5bae\",[],\", and from that, I generate UML diagrams of how that should work. I then review, accept changes, and create issues and tasks, then create a development report of any issues. After everything is complete, I will send the output to the PR review skill. In this way, I have full control over the project.”\",[1123],{\"_37\":1116,\"_23\":49,\"_439\":1124},\"https://en.wikipedia.org/wiki/Test-driven_development\",{\"_37\":1126,\"_23\":313,\"_314\":1127,\"_322\":1132,\"_324\":408},\"f051dcb87908\",[1128],{\"_37\":1129,\"_23\":318,\"_319\":1130,\"_174\":1131},\"4753a22209e7\",[],\"Accelerating AI Learning by Modeling Human Learning\",[],{\"_37\":1134,\"_23\":313,\"_314\":1135,\"_322\":1140,\"_324\":325},\"c08bceb94e27\",[1136],{\"_37\":1137,\"_23\":318,\"_319\":1138,\"_174\":1139},\"7cee6c784867\",[],\"Karaaslan suggests that his approach building skills started in much the same way he, himself, learns new things. “Before using AI, [to learn new development topics] I would read documentation, take notes, and learn from example projects. At first, I mimicked this process.”\",[],{\"_37\":1142,\"_23\":313,\"_314\":1143,\"_322\":1148,\"_324\":325},\"14c9c3315912\",[1144],{\"_37\":1145,\"_23\":318,\"_319\":1146,\"_174\":1147},\"1201808c6b7d\",[],\"The developer took a similar approach to eventually making Skill Seekers a project capable of ingesting data from a variety of different formats. “To implement how I learn things by watching videos into Skill Seekers, I would get code from the screen, collect timelines, and copy transcripts and screenshots into Skill Seekers. I was just mimicking how I learn without AI.”\",[],{\"_37\":1150,\"_23\":313,\"_314\":1151,\"_322\":1156,\"_324\":325},\"36482ac1dd9f\",[1152],{\"_37\":1153,\"_23\":318,\"_319\":1154,\"_174\":1155},\"20f514bf2178\",[],\"The developer notes that after the release of the project, he observed how others would use it, noting that users would want to review new additions. “People would want to adjust some stuff, add some stuff, remove some stuff. I wanted to automate the process, so I created a skill for that, and afterwards, we created the workflows, and gave the skill to that workflow. Exactly what we would do manually. It’s about mimicking, optimizing, and automating.”\",[],{\"_37\":1158,\"_23\":313,\"_314\":1159,\"_322\":1164,\"_324\":408},\"053310c32a04\",[1160],{\"_37\":1161,\"_23\":318,\"_319\":1162,\"_174\":1163},\"dd4beb9b89f7\",[],\"What Skills Add for Developers\",[],{\"_37\":1166,\"_23\":313,\"_314\":1167,\"_322\":1172,\"_324\":325},\"2a34dfe0e3ec\",[1168],{\"_37\":1169,\"_23\":318,\"_319\":1170,\"_174\":1171},\"170e2e448af6\",[],\"Karaaslan suggests a variety of development use cases have emerged from the project. “It started as a helper, but it has become a data management system for any kind of input that can be structured and reproduced, and with minimal cost because once you create a new skill config file, you can run that config again and again with a single command-line prompt, or just by telling Claude to do it directly via MCP.”\",[],{\"_37\":1174,\"_23\":313,\"_314\":1175,\"_322\":1180,\"_324\":325},\"60b04fe7b354\",[1176],{\"_37\":1177,\"_23\":318,\"_319\":1178,\"_174\":1179},\"947fd15a72bc\",[],\"The developer lists a variety of use cases, including an integration partner that uses the project to power its marketing and helper bots with skills that ping codebases to answer user questions about them. An academic AI research startup uses the project to scan new research papers for usable skills that the team can then use internally.\",[],{\"_37\":1182,\"_23\":313,\"_314\":1183,\"_322\":1188,\"_324\":325},\"3f29aaff6250\",[1184],{\"_37\":1185,\"_23\":318,\"_319\":1186,\"_174\":1187},\"98e9c7dcb932\",[],\"“Like I said, it's all just data that AI can understand. You can update to new versions easily and connect to your codebase in one click. So you can create your own tools, your own skills from your codebase. You can give it official documentation, your own implementation, examples, maybe some video, and it all gets merged into one big skill that includes everything.”\",[],{\"_37\":1190,\"_23\":313,\"_314\":1191,\"_322\":1196,\"_324\":325},\"7b39198592ae\",[1192],{\"_37\":1193,\"_23\":318,\"_319\":1194,\"_174\":1195},\"14803bd91b2b\",[],\"“After you create skills, you can add a workflow to create an agent and system calls. I use that feature often because you can control the context. It's actually a helpful tool for context engineering.”\",[],{\"_37\":1198,\"_23\":313,\"_314\":1199,\"_322\":1204,\"_324\":325},\"c20406bd4d1e\",[1200],{\"_37\":1201,\"_23\":318,\"_319\":1202,\"_174\":1203},\"d547b9b72703\",[],\"“In my workflows I have an automated skill that's triggered every time I push to the main repo. Every time I want to do design work I have a ‘design repos’ skill that I create with Skill Seekers focusing on the architecture and API. Whenever I run my agent skill, it automatically uploads that skill, and every time I run it, I know I have the latest version of my architecture as knowledge within the context inside the agent.”\",[],{\"_37\":1206,\"_23\":313,\"_314\":1207,\"_322\":1212,\"_324\":408},\"6c069a299096\",[1208],{\"_37\":1209,\"_23\":318,\"_319\":1210,\"_174\":1211},\"c6643ea19115\",[],\"How to Think About Skills for Beginners\",[],{\"_37\":1214,\"_23\":313,\"_314\":1215,\"_322\":1220,\"_324\":325},\"7c264860c3cd\",[1216],{\"_37\":1217,\"_23\":318,\"_319\":1218,\"_174\":1219},\"120914fe75d7\",[],\"“Let's say you’re not very fond of AI, and you just want to do your work as a developer. Developers always think in terms of architecture, how we do stuff, how we structure stuff. If you want to use some new back-end framework or a new language you’re not experienced with, you can stay focused on system architecture while Skill Seekers builds a skill that turns your agent into an expert on your tech stack. The best part is that Skill Seekers has an MCP, so you can just casually tell Claude to create skills for your project. No extra steps.”\",[],{\"_37\":1222,\"_23\":313,\"_314\":1223,\"_322\":1228,\"_324\":325},\"9fdec64484a8\",[1224],{\"_37\":1225,\"_23\":318,\"_319\":1226,\"_174\":1227},\"97e51aae6b72\",[],\"“Let’s say we wanted to start a new project and it's in an area we’re not familiar with. We might want to do some light experimentation before building because the cost of learning seems too high. Previously, we might avoid focusing too much on learning entirely new frameworks and languages. We would tend to just stay with what we know now.”\",[],{\"_37\":1230,\"_23\":313,\"_314\":1231,\"_322\":1236,\"_324\":325},\"dd44cdbcb585\",[1232],{\"_37\":1233,\"_23\":318,\"_319\":1234,\"_174\":1235},\"ce8639b148b0\",[],\"“You can tell Skill Seekers, ‘I want to do this project, I want to use this technology, I want to try this new framework, I want to try this new library,’ in casual, natural language. The MCP will trigger and generate the skill for the latest, top-of-the-line skill for you. You can just say, ‘I have an idea’ or ‘I want to try this functionality of this language or this library,’ and you start writing anything that you can test.”\",[],{\"_37\":1238,\"_23\":313,\"_314\":1239,\"_322\":1244,\"_324\":325},\"59cb372ad7f7\",[1240],{\"_37\":1241,\"_23\":318,\"_319\":1242,\"_174\":1243},\"906b345d5b64\",[],\"\\\"It's not too different from being an engineering lead. You say, 'I have a vision. I want to do this thing,' and it creates a team for you that knows that field and executes on your vision. You have a problem, you have a solution, you have a tool, you have a limitation, you have an output. This is always the same.\\\"\",[],{\"_37\":1246,\"_23\":313,\"_314\":1247,\"_322\":1252,\"_324\":408},\"7d1e3ba7cadf\",[1248],{\"_37\":1249,\"_23\":318,\"_319\":1250,\"_174\":1251},\"e084f1fb50fe\",[],\"Real-World Use Cases for Skills\",[],{\"_37\":1254,\"_23\":313,\"_314\":1255,\"_322\":1260,\"_324\":325},\"d68d42826b18\",[1256],{\"_37\":1257,\"_23\":318,\"_319\":1258,\"_174\":1259},\"35bd51943844\",[],\"\\\"The biggest use case I see is builders using skills to power support for their GitHub repos across the different language versions they need to maintain. Each version has its own skill, and every time they push an update, an automatic trigger refreshes the bot's knowledge. Doing that by hand is incredibly time-consuming.\\\"\",[],{\"_37\":1262,\"_23\":313,\"_314\":1263,\"_322\":1285,\"_324\":325},\"98cd664b0b07\",[1264,1268,1273,1277,1281],{\"_37\":1265,\"_23\":318,\"_319\":1266,\"_174\":1267},\"be95be2426a8\",[],\"Karaaslan's most-used example comes from inside his own company, where the team builds with \",{\"_37\":1269,\"_23\":318,\"_319\":1270,\"_174\":1272},\"05b1ae8efc88\",[1271],\"4fa581774301\",\"Unity\",{\"_37\":1274,\"_23\":318,\"_319\":1275,\"_174\":1276},\"66ef3336e6bd\",[],\". He started by setting up an internal skill marketplace so developers could share work with each other, then began authoring skill config files for the technologies the team uses day-to-day. It didn't take long to spot the bottleneck. \\\"Even with Skill Seekers doing the hard part, the surrounding loop was repetitive: Create a config, generate a skill, package it, and manually upload it to the marketplace. The friction was in everything \",{\"_37\":1278,\"_23\":318,\"_319\":1279,\"_174\":1280},\"51b1214abac9\",[508],\"around\",{\"_37\":1282,\"_23\":318,\"_319\":1283,\"_174\":1284},\"b66638f931de\",[],\" the generation step.\\\"\",[1286],{\"_37\":1271,\"_23\":49,\"_439\":1287},\"https://en.wikipedia.org/wiki/Unity_(game_engine)\",{\"_37\":1289,\"_23\":313,\"_314\":1290,\"_322\":1295,\"_324\":325},\"ad21335a0fad\",[1291],{\"_37\":1292,\"_23\":318,\"_319\":1293,\"_174\":1294},\"d25c294230ef\",[],\"The developer explains his process of layering in fixes, one at a time. \\\"First, I added a shared config repo so anyone on the team could publish or pull a config through their own Skill Seekers install. Then, I added packaging and uploading to the marketplace as native functionality inside Skill Seekers itself. The flow worked end-to-end, but I hit a new wall. The volume of data I was sending Claude during the local enhancement step burned through my usage almost immediately.\\\"\",[],{\"_37\":1297,\"_23\":313,\"_314\":1298,\"_322\":1303,\"_324\":325},\"20a20aba9098\",[1299],{\"_37\":1300,\"_23\":318,\"_319\":1301,\"_174\":1302},\"1756cff382b4\",[],\"The fix was to stop tying Skill Seekers to any single model. \\\"I made it agent-, CLI-, and LLM-agnostic. Now I trigger the whole pipeline from the command line: Scrape, restructure, combine sources, run the local AI enhancement step on whichever model I want, package, and push straight to our internal marketplace. I tend to use OpenCode with Kimi for the enhancement step. I tend to get more usage with it, and in my own tests, the skill quality came out on par with Claude.\\\"\",[],{\"_37\":1305,\"_23\":313,\"_314\":1306,\"_322\":1311,\"_324\":325},\"66b38b1b1189\",[1307],{\"_37\":1308,\"_23\":318,\"_319\":1309,\"_174\":1310},\"0b2ed57a9d74\",[],\"That decoupling is also where Karaaslan sees the project's longer-term direction. \\\"My vision for Skill Seekers is for it to become a universal data layer for AI systems. Model-agnostic, source-agnostic, pipeline-friendly. Whatever the input, whatever the agent on the other end, the skill is the same shape. That's the point.\\\"\",[],{\"_37\":1313,\"_23\":313,\"_314\":1314,\"_322\":1319,\"_324\":325},\"9d01d9f2531d\",[1315],{\"_37\":1316,\"_23\":318,\"_319\":1317,\"_174\":1318},\"5a4f419b23ed\",[],\"\\n\",[],[1321],{\"_37\":1322,\"_23\":313,\"_314\":1323,\"_322\":1328,\"_324\":325},\"299596cf3478\",[1324],{\"_37\":1325,\"_23\":318,\"_319\":1326,\"_174\":1327},\"fcdd5d183d68\",[],\"Developer Yusuf Karaaslan explains how he ended up designing a system that translates a variety of data into AI skills.\",[],[1330,1333,1336],{\"_37\":1331,\"_116\":1332,\"_23\":118},\"tHaSHEg5WYHt\",\"8905ca81-07ca-43e9-93dd-488c95cb3e63\",{\"_37\":1334,\"_116\":1335,\"_23\":118},\"KnbBSvoF6FOr\",\"13c1b831-4d81-4d3e-a5ad-97fc16d01109\",{\"_37\":1337,\"_116\":1338,\"_23\":118},\"mSmtizdTuuKY\",\"359105c7-1b7e-4ff8-9c6a-bf55c78db533\",\"2026-05-12T16:42:00.000Z\",{\"_147\":1341,\"_23\":295,\"_28\":1327,\"_298\":196,\"_299\":196,\"_32\":1344},{\"_23\":149,\"_114\":1342},{\"_116\":1343,\"_23\":118},\"image-c1d9aa33991bb4aa539fa42b8cc3f6d80638ba0d-1200x630-png\",\"Translating Data into AI Skills\",{\"_23\":366,\"_368\":1346},\"translating-data-into-ai-skills\",{\"_11\":1348,\"_13\":1349,\"_14\":1350,\"_23\":278,\"_24\":1351,\"_281\":1352,\"_393\":1357,\"_398\":1358,\"_891\":1907,\"_901\":1916,\"_912\":1926,\"_295\":1927,\"_366\":1928,\"_920\":1930,\"_943\":944,\"_32\":1942},\"2026-07-15T22:53:15Z\",\"05705342-f90c-45e4-9455-efa317ecd74c\",\"g4rJ04viFnWjgGFaBKUVP9\",\"2026-07-21T18:06:11Z\",[1353],{\"_37\":1354,\"_23\":285,\"_286\":1355,\"_300\":1356},\"4620291c2def\",{\"_116\":289,\"_23\":118},{\"_116\":392,\"_23\":118},[],[1359,1367,1408,1412,1427,1435,1454,1462,1524,1540,1564,1572,1580,1588,1596,1612,1631,1639,1648,1656,1664,1683,1691,1699,1707,1715,1723,1731,1739,1755,1796,1804,1812,1820,1828,1836,1844,1852,1867,1875,1883,1891,1899],{\"_37\":1360,\"_23\":313,\"_314\":1361,\"_322\":1366,\"_324\":408},\"a4dd381f0a68\",[1362],{\"_37\":1363,\"_23\":318,\"_319\":1364,\"_174\":1365},\"e6a4b1036650\",[],\"Performance Engineering: Part Brute-Force, Part Hard Decisions?\",[],{\"_37\":1368,\"_23\":313,\"_314\":1369,\"_322\":1401,\"_324\":325},\"6b55dc2b123d\",[1370,1374,1379,1383,1388,1392,1397],{\"_37\":1371,\"_23\":318,\"_319\":1372,\"_174\":1373},\"4e7c49a1f197\",[],\"Software startups often launch quickly to beat competitors and gather early customer feedback. At that stage, performance should not necessarily be the priority, founder \",{\"_37\":1375,\"_23\":318,\"_319\":1376,\"_174\":1378},\"878b895278dc\",[1377],\"e7b6a348b24d\",\"Tomás Senart\",{\"_37\":1380,\"_23\":318,\"_319\":1381,\"_174\":1382},\"c0bbd67a05eb\",[],\" argues. But once a product is exposed to scale, or latency and infrastructure costs become material, the economics change and teams need a repeatable way to keep performance from drifting. The creator of the popular Go HTTP load-testing tool \",{\"_37\":1384,\"_23\":318,\"_319\":1385,\"_174\":1387},\"8035a870ffe7\",[1386],\"5580234d1c49\",\"Vegeta\",{\"_37\":1389,\"_23\":318,\"_319\":1390,\"_174\":1391},\"d2a81850e049\",[],\" has carried those lessons into his startup \",{\"_37\":1393,\"_23\":318,\"_319\":1394,\"_174\":1396},\"b274bfb3feec\",[1395],\"84ac9a50d23d\",\"Perfloop\",{\"_37\":1398,\"_23\":318,\"_319\":1399,\"_174\":1400},\"0887208e2c5b\",[],\", and shares them here.\",[1402,1404,1406],{\"_37\":1377,\"_23\":49,\"_439\":1403},\"https://www.linkedin.com/in/tsenart/\",{\"_37\":1386,\"_23\":49,\"_439\":1405},\"https://github.com/tsenart/vegeta\",{\"_37\":1395,\"_23\":49,\"_439\":1407},\"https://perfloop.ai\",{\"_37\":1409,\"_23\":149,\"_114\":1410},\"48a511be4425\",{\"_116\":1411,\"_23\":118},\"image-f3af6263901812ce5c45efe91414caaa22705694-3840x1843-jpg\",{\"_37\":1413,\"_23\":313,\"_314\":1414,\"_322\":1424,\"_324\":325},\"3a29ea332842\",[1415,1419],{\"_37\":1416,\"_23\":318,\"_319\":1417,\"_174\":1418},\"6bb710a15cda\",[508],\"Tomás Senart discusses his learnings in Go interfaces at GopherCon 2015. Image courtesy \",{\"_37\":1420,\"_23\":318,\"_319\":1421,\"_174\":1423},\"cfd99cdab938\",[1422,508],\"6d8313316171\",\"Gopher Academy\",[1425],{\"_37\":1422,\"_23\":49,\"_439\":1426},\"https://www.youtube.com/watch?v=xyDkyFjzFVc\",{\"_37\":1428,\"_23\":313,\"_314\":1429,\"_322\":1434,\"_324\":408},\"0af3ab4dba7d\",[1430],{\"_37\":1431,\"_23\":318,\"_319\":1432,\"_174\":1433},\"fffe68f9eae2\",[],\"From Load Testing to Go\",[],{\"_37\":1436,\"_23\":313,\"_314\":1437,\"_322\":1451,\"_324\":325},\"6350ff8084c4\",[1438,1442,1447],{\"_37\":1439,\"_23\":318,\"_319\":1440,\"_174\":1441},\"a1f09e622ab3\",[],\"As an engineer at SoundCloud, Senart had been tasked with replacing a significant chunk of a legacy system that needed to be loaded and tested. “At that time, the only thing available that was not a [heavyweight] Java ecosystem tool was \",{\"_37\":1443,\"_23\":318,\"_319\":1444,\"_174\":1446},\"ae381f506219\",[1445],\"b618c96c6b8b\",\"Apache Benchmark\",{\"_37\":1448,\"_23\":318,\"_319\":1449,\"_174\":1450},\"dc39afa33854\",[],\", which had an array of usability issues and functionality limitations. We had a concrete problem and a culture of building things from scratch. Those two things together led to Vegeta.”\",[1452],{\"_37\":1445,\"_23\":49,\"_439\":1453},\"https://en.wikipedia.org/wiki/ApacheBench\",{\"_37\":1455,\"_23\":313,\"_314\":1456,\"_322\":1461,\"_324\":325},\"25c5ca8a58e4\",[1457],{\"_37\":1458,\"_23\":318,\"_319\":1459,\"_174\":1460},\"67a151e612d4\",[],\"While the project has become popular in the Golang community, the founder clarifies that there was more to its success than the inherent speed and transparency of the language itself. “I think you can do performance engineering in any language ecosystem. With Go, because concurrency was such a core part of the design of the language, it made building a load-testing tool very natural at the time.”\",[],{\"_37\":1463,\"_23\":313,\"_314\":1464,\"_322\":1519,\"_324\":325},\"e5a8865687f9\",[1465,1469,1473,1477,1481,1485,1490,1494,1499,1503,1507,1511,1515],{\"_37\":1466,\"_23\":318,\"_319\":1467,\"_174\":1468},\"619173514fbe\",[],\"“But you can build a load-testing tool in a way that will ‘lie to you’ in the results. So it wasn't just about building something \",{\"_37\":1470,\"_23\":318,\"_319\":1471,\"_174\":1472},\"e80707fa5053\",[508],\"quick\",{\"_37\":1474,\"_23\":318,\"_319\":1475,\"_174\":1476},\"a971fcca11b6\",[],\", it was about building something \",{\"_37\":1478,\"_23\":318,\"_319\":1479,\"_174\":1480},\"7be9c958a9af\",[508],\"correct\",{\"_37\":1482,\"_23\":318,\"_319\":1483,\"_174\":1484},\"c1a4323f18c1\",[],\". Azul Systems CTO \",{\"_37\":1486,\"_23\":318,\"_319\":1487,\"_174\":1489},\"1fbeac92fe7c\",[1488],\"f3a678ca4971\",\"Gil Tene\",{\"_37\":1491,\"_23\":318,\"_319\":1492,\"_174\":1493},\"acd03976226a\",[],\" is famous in the performance engineering world for many things, including the concept of \",{\"_37\":1495,\"_23\":318,\"_319\":1496,\"_174\":1498},\"69062ea847c5\",[1497,508],\"35c4921dad59\",\"coordinated omission\",{\"_37\":1500,\"_23\":318,\"_319\":1501,\"_174\":1502},\"a5521ff0a552\",[],\": A phenomenon in load testing (and many tools fall into this trap) where the client that is sending the requests \",{\"_37\":1504,\"_23\":318,\"_319\":1505,\"_174\":1506},\"2f3ff173870b\",[508],\"paces itself\",{\"_37\":1508,\"_23\":318,\"_319\":1509,\"_174\":1510},\"252e589213b2\",[],\" according to \",{\"_37\":1512,\"_23\":318,\"_319\":1513,\"_174\":1514},\"e30a30c5a11a\",[508],\"how the server is doing\",{\"_37\":1516,\"_23\":318,\"_319\":1517,\"_174\":1518},\"da4862c0a9d8\",[],\".”\",[1520,1522],{\"_37\":1488,\"_23\":49,\"_439\":1521},\"https://www.linkedin.com/in/giltene/k\",{\"_37\":1497,\"_23\":49,\"_439\":1523},\"https://www.youtube.com/watch?v=6Rs0p3mPNr0\u0026t=19s\",{\"_37\":1525,\"_23\":313,\"_314\":1526,\"_322\":1539,\"_324\":325},\"10eeb80dd561\",[1527,1531,1535],{\"_37\":1528,\"_23\":318,\"_319\":1529,\"_174\":1530},\"98b3035e7803\",[],\"“That is a bad way to test the system, because you're basically stopping to hammer the server at the exact point where you'd get the [indication] that it's not doing well. So you need to keep the rate constant. That was a key design decision at the time to build Vegeta in this way: To \",{\"_37\":1532,\"_23\":318,\"_319\":1533,\"_174\":1534},\"a8720b9fe5c0\",[508],\"not\",{\"_37\":1536,\"_23\":318,\"_319\":1537,\"_174\":1538},\"98d1a82bc4e6\",[],\" suffer from that problem.”\",[],{\"_37\":1541,\"_23\":313,\"_314\":1542,\"_322\":1563,\"_324\":325},\"32f1b6f27da0\",[1543,1547,1551,1555,1559],{\"_37\":1544,\"_23\":318,\"_319\":1545,\"_174\":1546},\"7e874dacbb3f\",[],\"The founder suggests that the most important lesson from his earlier work wasn’t that performance data is \",{\"_37\":1548,\"_23\":318,\"_319\":1549,\"_174\":1550},\"d92d6a26dfe1\",[508],\"lacking\",{\"_37\":1552,\"_23\":318,\"_319\":1553,\"_174\":1554},\"3d2e96a4d7e6\",[],\", as most teams have dashboards full of telemetry and a benchmark they’ve run at least once. “The problem is that the benchmark you ran \",{\"_37\":1556,\"_23\":318,\"_319\":1557,\"_174\":1558},\"713acd7a8f31\",[508],\"once\",{\"_37\":1560,\"_23\":318,\"_319\":1561,\"_174\":1562},\"4b7cfbd50efd\",[],\" is already stale. Performance is not something you set and keep, it is an envelope you have to hold against constant change.”\",[],{\"_37\":1565,\"_23\":313,\"_314\":1566,\"_322\":1571,\"_324\":325},\"e8ab65fddf6f\",[1567],{\"_37\":1568,\"_23\":318,\"_319\":1569,\"_174\":1570},\"be64909a2de9\",[],\"“Every code change, dependency bump, traffic shift, and architectural change nudges it, so the loop from measurement to a verified improvement has to run again and again just to stay in place.” Senart explains that the measurement process is inherently slow, which makes it easy to deprioritize and fall by the wayside.\",[],{\"_37\":1573,\"_23\":313,\"_314\":1574,\"_322\":1579,\"_324\":325},\"6aab59def7f8\",[1575],{\"_37\":1576,\"_23\":318,\"_319\":1577,\"_174\":1578},\"f081372f7995\",[],\"His startup Perfloop closes the second half of the loop: After measuring, Perfloop assesses how systems behave under load, across the entire loop. “It finds the performance opportunities, proposes changes, and proves whether they actually moved the needle, with the numbers to back it up.”\",[],{\"_37\":1581,\"_23\":313,\"_314\":1582,\"_322\":1587,\"_324\":325},\"5a5229941c42\",[1583],{\"_37\":1584,\"_23\":318,\"_319\":1585,\"_174\":1586},\"31f0eb539ffc\",[],\"Providing proof, the founder argues, validates automation across the rest of the process, rather than forcing teams to take things on faith: An important nuance as benchmarking loops need to be run far more often than can be done manually. “The system carries the ‘relentless’ part, the finding and the proving, while the person stays in charge of what is worth holding and what actually ships.”\",[],{\"_37\":1589,\"_23\":313,\"_314\":1590,\"_322\":1595,\"_324\":408},\"5c1a8f426075\",[1591],{\"_37\":1592,\"_23\":318,\"_319\":1593,\"_174\":1594},\"9765ff86a5e9\",[],\"How Much Should Tech Leaders Prioritize Benchmarks, Really?\",[],{\"_37\":1597,\"_23\":313,\"_314\":1598,\"_322\":1611,\"_324\":325},\"eb78531892d8\",[1599,1603,1607],{\"_37\":1600,\"_23\":318,\"_319\":1601,\"_174\":1602},\"d1f7483e06d4\",[],\"Addressing the excitement around AI evaluations, the founder notes that the most popular AI benchmarks can and do become contaminated. “That's why people started creating private eval sets that they run so they're not gameable. That’s one approach, but it has its own problems, because there is no transparency into the \",{\"_37\":1604,\"_23\":318,\"_319\":1605,\"_174\":1606},\"c055f7d34db5\",[508],\"quality\",{\"_37\":1608,\"_23\":318,\"_319\":1609,\"_174\":1610},\"7a828099ba8d\",[],\" of those benchmarks. It's a moving target, honestly: How to have transparency while not being gameable or having evals end up in the training data of models themselves.”\",[],{\"_37\":1613,\"_23\":313,\"_314\":1614,\"_322\":1628,\"_324\":325},\"f4c4c7a9d25b\",[1615,1619,1624],{\"_37\":1616,\"_23\":318,\"_319\":1617,\"_174\":1618},\"44b0492f53e2\",[],\"Senart draws a line between viral LLM benchmarks like \",{\"_37\":1620,\"_23\":318,\"_319\":1621,\"_174\":1623},\"096d06d031a8\",[1622],\"92e68e524ba1\",\"Deep-SWE\",{\"_37\":1625,\"_23\":318,\"_319\":1626,\"_174\":1627},\"9c59af093449\",[],\" and benchmarks that teams specifically use internally to build their own products. “If you're building these capabilities in your company and in your product, the only thing that matters is your evals. It's not going to be the public coding agent evals for the coding models that matter the most.”\",[1629],{\"_37\":1622,\"_23\":49,\"_439\":1630},\"https://deepswe.datacurve.ai\",{\"_37\":1632,\"_23\":313,\"_314\":1633,\"_322\":1638,\"_324\":325},\"28a4b6f2be72\",[1634],{\"_37\":1635,\"_23\":318,\"_319\":1636,\"_174\":1637},\"95798f00fb94\",[],\"“The benchmarks that last are the unglamorous, private ones tied to an outcome you actually care about, because there is no incentive to game a benchmark whose only audience is you. Public benchmarks help you decide what to test. Your own codebase tells you what is true.”\",[],{\"_37\":1640,\"_23\":313,\"_314\":1641,\"_322\":1646,\"_324\":1647},\"a5389682f039\",[1642],{\"_37\":1643,\"_23\":318,\"_319\":1644,\"_174\":1645},\"c37664ea11b9\",[],\"\\\"The only thing that matters is your evals. The benchmarks that last are the unglamorous, private ones tied to an outcome you actually care about, because there is no incentive to game a benchmark whose only audience is you. Your own codebase tells you what is true.” -Tomás Senart, Founder/Perfloop\",[],\"blockquote\",{\"_37\":1649,\"_23\":313,\"_314\":1650,\"_322\":1655,\"_324\":408},\"876394a905c4\",[1651],{\"_37\":1652,\"_23\":318,\"_319\":1653,\"_174\":1654},\"0f91aee922f8\",[],\"Performance Engineering Lessons for Founders\",[],{\"_37\":1657,\"_23\":313,\"_314\":1658,\"_322\":1663,\"_324\":325},\"83c108e99621\",[1659],{\"_37\":1660,\"_23\":318,\"_319\":1661,\"_174\":1662},\"0c9625d1f1ab\",[],\"Senart suggests that the adoption of agentic coding agents is fundamentally changing software, and not necessarily for the better. “First, agents now write a large and growing share of code, with a very wide range of outcomes. Many teams are moving faster than their ability to review every line, so they are shipping performance slop: Inefficiencies and waste that no human deliberately introduced and no one is specifically watching for.”\",[],{\"_37\":1665,\"_23\":313,\"_314\":1666,\"_322\":1680,\"_324\":325},\"118165504c61\",[1667,1671,1676],{\"_37\":1668,\"_23\":318,\"_319\":1669,\"_174\":1670},\"c3bdc4829b49\",[],\"“Second, the load on production systems is changing shape. It is not only human users anymore. Agents call services at machine speed and volume, so the same systems get hammered far harder than they were designed for.” The founder observes that \",{\"_37\":1672,\"_23\":318,\"_319\":1673,\"_174\":1675},\"47238b41abf9\",[1674],\"402934988603\",\"bot Web traffic has surpassed human traffic\",{\"_37\":1677,\"_23\":318,\"_319\":1678,\"_174\":1679},\"038694828348\",[],\", and Web-based infrastructure is only the first beachhead to be pounded on by agents.\",[1681],{\"_37\":1674,\"_23\":49,\"_439\":1682},\"https://www.nbcnews.com/tech/tech-news/bot-web-traffic-overtaken-human-web-traffic-data-shows-rcna348522\",{\"_37\":1684,\"_23\":313,\"_314\":1685,\"_322\":1690,\"_324\":325},\"f2138f0b862e\",[1686],{\"_37\":1687,\"_23\":318,\"_319\":1688,\"_174\":1689},\"4699fd015f3f\",[],\"Commerce is another rapidly-growing space for agentic workflows for retail and checkout processes, and enterprise services appear to be the next frontier. The founder argues that agentic traffic is no longer a hypothetical future condition. It’s an inevitable thing that will happen, and make performance even more crucial.\",[],{\"_37\":1692,\"_23\":313,\"_314\":1693,\"_322\":1698,\"_324\":325},\"9d269ae62ef1\",[1694],{\"_37\":1695,\"_23\":318,\"_319\":1696,\"_174\":1697},\"c94788290786\",[],\"In addition, it’s no longer feasible to wait to be ‘bailed out’ by eventually better hardware. “Now GPUs, memory, and compute are expensive and staying that way, so inefficiency shows up directly as a cost on your bill.”\",[],{\"_37\":1700,\"_23\":313,\"_314\":1701,\"_322\":1706,\"_324\":325},\"f90f5beaa2fd\",[1702],{\"_37\":1703,\"_23\":318,\"_319\":1704,\"_174\":1705},\"dfd1ac8f0ab4\",[],\"The above factors lead to more performance debt that accrues quicker on infrastructure taking heavier workloads. “What is missing is a way to keep up. Occasional expert attention or a periodic consultant audit cannot match how fast the debt now accrues. That is the gap Perfloop is built to close.”\",[],{\"_37\":1708,\"_23\":313,\"_314\":1709,\"_322\":1714,\"_324\":325},\"8852aa099607\",[1710],{\"_37\":1711,\"_23\":318,\"_319\":1712,\"_174\":1713},\"d746e722093c\",[],\"The founder clarifies that while you can use agentic systems to support performance engineering, AI doesn’t nullify the fundamentals. “You still have resources, you have things that drag on them, and you’ve got to make the most of them. You need to unpack that, one layer at a time, and be methodical. That's what performance engineering is.”\",[],{\"_37\":1716,\"_23\":313,\"_314\":1717,\"_322\":1722,\"_324\":325},\"d3ea0bba3058\",[1718],{\"_37\":1719,\"_23\":318,\"_319\":1720,\"_174\":1721},\"65f23abdfb30\",[],\"“You still create hypotheses, you go test them, and you look at the hard data. It’s just that the experiments can go faster now. You have agents which, with the right harness, can do that kind of work rigorously and not just guessing and going in the wrong direction, then you can parallelize so many of these hypotheses for testing.”\",[],{\"_37\":1724,\"_23\":313,\"_314\":1725,\"_322\":1730,\"_324\":325},\"f161dafd0bca\",[1726],{\"_37\":1727,\"_23\":318,\"_319\":1728,\"_174\":1729},\"4554c83a6eca\",[],\"Senart offers two final pieces of advice for founders. “First: Do not pay for performance you do not need yet; if nobody uses your product, its performance does not matter. Second: The moment you are exposed to scale, the economics change in your favor. Performance work is methodical science, and the old constraint was that you could only afford to test the top one or two ideas. With the right harness you can now run many of those experiments in parallel, with rigor, and test the other ten you used to ignore.”\",[],{\"_37\":1732,\"_23\":313,\"_314\":1733,\"_322\":1738,\"_324\":408},\"57bbc9baf3e9\",[1734],{\"_37\":1735,\"_23\":318,\"_319\":1736,\"_174\":1737},\"4a9a526c2974\",[],\"Getting Performance Engineering into Enterprise Customers’ Hands\",[],{\"_37\":1740,\"_23\":313,\"_314\":1741,\"_322\":1754,\"_324\":325},\"fa3c385c950e\",[1742,1746,1750],{\"_37\":1743,\"_23\":318,\"_319\":1744,\"_174\":1745},\"d0cd28533dac\",[],\"Senart concedes that his startup’s approach to \",{\"_37\":1747,\"_23\":318,\"_319\":1748,\"_174\":1749},\"7df8713b8424\",[508],\"automating \",{\"_37\":1751,\"_23\":318,\"_319\":1752,\"_174\":1753},\"00808c223c36\",[],\"performance engineering isn’t necessarily mainstream among enterprise engineering orgs yet, due to a number of factors. “The first is trust, which is the hard one. You are asking a large organization to let an autonomous system reason about and propose changes to code that runs their business. They will not do that on a verbal promise.\",[],{\"_37\":1756,\"_23\":313,\"_314\":1757,\"_322\":1789,\"_324\":325},\"7ab4e5c9bade\",[1758,1762,1767,1771,1776,1780,1785],{\"_37\":1759,\"_23\":318,\"_319\":1760,\"_174\":1761},\"8afeb9a04ea2\",[],\"“They need proof that the system finds real problems and that its proposed changes actually work, and they need that proof to come from somewhere they cannot dismiss. This is exactly why we are doing the work in \",{\"_37\":1763,\"_23\":318,\"_319\":1764,\"_174\":1766},\"0b23f2b5ef55\",[1765],\"023f344d3abc\",\"open source, in public\",{\"_37\":1768,\"_23\":318,\"_319\":1769,\"_174\":1770},\"5dfa889bcde1\",[],\".” The founder notes that the maintainer of the widely-used Go library \",{\"_37\":1772,\"_23\":318,\"_319\":1773,\"_174\":1775},\"75f8ece5f8f6\",[1774],\"4a41e7ef1d81\",\"parquet-go\",{\"_37\":1777,\"_23\":318,\"_319\":1778,\"_174\":1779},\"ddb4559759a8\",[],\" has \",{\"_37\":1781,\"_23\":318,\"_319\":1782,\"_174\":1784},\"3809cf474701\",[1783],\"5e42da5cf296\",\"publicly agreed\",{\"_37\":1786,\"_23\":318,\"_319\":1787,\"_174\":1788},\"9345986d8863\",[],\" with his hypothesis that closing performance debt will come down to utilizing the kind of concurrent optimizations he’s building today.\",[1790,1792,1794],{\"_37\":1765,\"_23\":49,\"_439\":1791},\"https://app.perfloop.ai/t/oss/roi\",{\"_37\":1774,\"_23\":49,\"_439\":1793},\"https://github.com/parquet-go/parquet-go\",{\"_37\":1783,\"_23\":49,\"_439\":1795},\"https://x.com/__Achille__/status/2068578598521594131?s=20\",{\"_37\":1797,\"_23\":313,\"_314\":1798,\"_322\":1803,\"_324\":325},\"db3ee98b4732\",[1799],{\"_37\":1800,\"_23\":318,\"_319\":1801,\"_174\":1802},\"c4c16313a7d9\",[],\"“The second blocker is the usual enterprise gate: Compliance, security review, governance. There is nothing special here, it is just work you have to do to be allowed in the building. We are starting SOC 2 now for that reason.”\",[],{\"_37\":1805,\"_23\":313,\"_314\":1806,\"_322\":1811,\"_324\":325},\"a54d47547e16\",[1807],{\"_37\":1808,\"_23\":318,\"_319\":1809,\"_174\":1810},\"e0ef4e839b26\",[],\"“The third blocker is not so much a blocker as it is inertia: Simply the reason to bother to do anything. Enterprises are actually a strong fit, because their scale and internal complexity produce a lot of recoverable inefficiency. In a large system, the performance you have left on the table is real money on the bill. Once trust and compliance are cleared, that recoverable cost is a very concrete reason to engage.”\",[],{\"_37\":1813,\"_23\":313,\"_314\":1814,\"_322\":1819,\"_324\":325},\"818c82cb1709\",[1815],{\"_37\":1816,\"_23\":318,\"_319\":1817,\"_174\":1818},\"a11e40641afa\",[],\"“What it takes [to get into enterprises] is portable, verifiable proof on code that looks like theirs, compliance work done honestly, and a value story told in the language they care about, which is cost and risk, not benchmarks for their own sake. For us, this is the route, not the first stop. The immediate motion is the public open-source work and self-serve adoption. The enterprise conversation opens once that public credibility is built.”\",[],{\"_37\":1821,\"_23\":313,\"_314\":1822,\"_322\":1827,\"_324\":1647},\"58407b0a1fc2\",[1823],{\"_37\":1824,\"_23\":318,\"_319\":1825,\"_174\":1826},\"9b3b17aee998\",[],\"\\\"[For enterprises], the first [blocker] is trust, which is the hard one. You are asking a large organization to let an autonomous system reason about and propose changes to code that runs their business. They will not do that on a verbal promise. The second blocker is the usual enterprise gate: Compliance, security review, governance. The third blocker is not so much a blocker as it is inertia: Simply the reason to bother to do anything.”\",[],{\"_37\":1829,\"_23\":313,\"_314\":1830,\"_322\":1835,\"_324\":408},\"9a70b4ad9aea\",[1831],{\"_37\":1832,\"_23\":318,\"_319\":1833,\"_174\":1834},\"78b303d327ce\",[],\"What the Future of Performance Engineering Looks Like\",[],{\"_37\":1837,\"_23\":313,\"_314\":1838,\"_322\":1843,\"_324\":325},\"6774c3f9c853\",[1839],{\"_37\":1840,\"_23\":318,\"_319\":1841,\"_174\":1842},\"c73730ce1a34\",[],\"Senart understands the current excitement about AI agents in software, but isn’t sure that performance engineering will simply become another agentic guardrail. “Guardrails are part of it, but the real question is where the human sits on the spectrum between specifying everything up front and steering every step (and both ends fail).”\",[],{\"_37\":1845,\"_23\":313,\"_314\":1846,\"_322\":1851,\"_324\":325},\"736ea2f89f04\",[1847],{\"_37\":1848,\"_23\":318,\"_319\":1849,\"_174\":1850},\"e5b0adca6d5f\",[],\"The founder offers the two extremes: The ‘dark factory’ that is prompted once while humans step back, without human judgement or context along the way. The other is today’s common practice of endlessly babysitting coding agents edit by edit, which preserves human judgement but can’t scale, turning senior engineers into bottlenecks.\",[],{\"_37\":1853,\"_23\":313,\"_314\":1854,\"_322\":1866,\"_324\":325},\"dfd34744e0b2\",[1855,1859,1863],{\"_37\":1856,\"_23\":318,\"_319\":1857,\"_174\":1858},\"55d49b92e2b4\",[],\"“What actually amplifies a senior engineer is in between, and it is the harder thing to build. The human stays on the decisions that carry judgment, which tradeoffs to accept and what is worth holding, while the system runs the steps underneath and surfaces only the calls that need a person. Present where judgment is required, absent everywhere else. Not steering more, steering \",{\"_37\":1860,\"_23\":318,\"_319\":1861,\"_174\":1862},\"b0410682ed1e\",[508],\"higher\",{\"_37\":1864,\"_23\":318,\"_319\":1865,\"_174\":1518},\"4cade10826e2\",[],[],{\"_37\":1868,\"_23\":313,\"_314\":1869,\"_322\":1874,\"_324\":325},\"1f895ba22398\",[1870],{\"_37\":1871,\"_23\":318,\"_319\":1872,\"_174\":1873},\"4593e958af10\",[],\"“Performance needs exactly this, because the target keeps moving. You cannot freeze the judgment into an up-front spec when the envelope shifts with every change, but you also cannot re-derive every step by hand each time. So you want a system that does the relentless step-level work and pulls the human in precisely when a real tradeoff is on the table.”\",[],{\"_37\":1876,\"_23\":313,\"_314\":1877,\"_322\":1882,\"_324\":325},\"85917623cc1a\",[1878],{\"_37\":1879,\"_23\":318,\"_319\":1880,\"_174\":1881},\"b71d478d831e\",[],\"“That is what Perfloop is built for. The whole loop runs as an MCP server, so the pieces are callable on their own: A person or an agent can point it at one slice of code, or hand it a rough hunch about what might be slow and get back a benchmarked case. Autonomy is a dial the customer controls, from scoping, to directing, to reviewing, to fully delegating, which is really just a way of choosing how high on that spectrum you sit.”\",[],{\"_37\":1884,\"_23\":313,\"_314\":1885,\"_322\":1890,\"_324\":325},\"69844ce85a03\",[1886],{\"_37\":1887,\"_23\":318,\"_319\":1888,\"_174\":1889},\"ca445a8b8cd2\",[],\"“My honest bet is most teams land at ‘direct’ or ‘review’ for anything that matters: Let the system carry the steps and the proof, but keep their hands on the decisions that need judgment. So performance engineering won’t shrink into a guardrail. It will remain a loop that humans steer from above, with far more reach than before.”\",[],{\"_37\":1892,\"_23\":313,\"_314\":1893,\"_322\":1898,\"_324\":325},\"44c6632fc6cd\",[1894],{\"_37\":1895,\"_23\":318,\"_319\":1896,\"_174\":1897},\"4738e59bec97\",[],\"The founder notes two additional factors that will affect future software projects. “First, the set of teams exposed to scale is expanding fast, because agent traffic drags services that used to be ‘safe’ into the deep end. Second, for the teams already at scale, the economics changed underneath them: The debt piles up faster, because agents write more code with less scrutiny per line, and it no longer depreciates the way it used to, because you can no longer count on cheaper, faster hardware to absorb it.”\",[],{\"_37\":1900,\"_23\":313,\"_314\":1901,\"_322\":1906,\"_324\":325},\"42705e1fddd8\",[1902],{\"_37\":1903,\"_23\":318,\"_319\":1904,\"_174\":1905},\"8b4b8729c0a3\",[],\"“Performance debt that used to quietly depreciate now accrues interest. When a problem goes from occasional to continuous, how you handle it has to change too. A once-a-year expert audit cannot keep up with debt that compounds. Perfloop is my bet on what replaces it: The rigor a great performance engineer brings, available to any team that needs it, with the human keeping the judgment and the system carrying the grind. The open-source work is where we are proving it in public, one verifiable win at a time.”\",[],[1908],{\"_37\":1909,\"_23\":313,\"_314\":1910,\"_322\":1915,\"_324\":325},\"881a0fc135fd\",[1911],{\"_37\":1912,\"_23\":318,\"_319\":1913,\"_174\":1914},\"453c407f68e5\",[],\"Founder Tomás Senart explains why the future of performance engineering will be AI doing the heavy lifting while humans make strategic calls.\",[],[1917,1920,1923],{\"_37\":1918,\"_116\":1919,\"_23\":118},\"X9mXpUXBqFgT\",\"321717a4-841b-4ccd-85d7-88d8971bb67a\",{\"_37\":1921,\"_116\":1922,\"_23\":118},\"Yfou29C6jhXl\",\"ad74ba93-b61f-443e-aa32-cfe709cbc7cf\",{\"_37\":1924,\"_116\":1925,\"_23\":118},\"sA62QdrBxLxp\",\"8a9f6d6c-8027-4b4d-8dbd-f93dfbce3216\",\"2026-07-16T15:47:00.000Z\",{\"_23\":295,\"_28\":1914,\"_298\":196,\"_299\":196},{\"_23\":366,\"_368\":1929},\"automating-performance-engineering-with-ai-and-humans-in-the-loop\",[1931,1934,1936,1939],{\"_37\":1932,\"_116\":1933,\"_23\":118},\"39a20f4c2cd1\",\"kFpYkfzWCiYI8xC60oV6DE\",{\"_37\":1935,\"_116\":927,\"_23\":118},\"ca063cdaae51\",{\"_37\":1937,\"_116\":1938,\"_23\":118},\"9a25c709ab67\",\"sGpx8ZAtjp3y9Udp02ENIG\",{\"_37\":1940,\"_116\":1941,\"_23\":118},\"49fece3e455b\",\"recxMLtjPJN8xUgWn\",\"Will AI Manage Software Performance with Humans in the Loop?\",[1944],{\"_11\":1945,\"_13\":1946,\"_14\":1947,\"_16\":1948,\"_23\":1951,\"_24\":1952,\"_1953\":1954,\"_1985\":171,\"_28\":1986,\"_149\":1987,\"_326\":1990,\"_295\":2047,\"_366\":2051,\"_32\":2053},\"2024-01-08T19:47:09Z\",\"9ab212d3-fdb7-4efc-b7e8-1a50b8988f2b\",\"hcyMVZIajME77768hd6qyg\",{\"_18\":1949},{\"_20\":1946,\"_21\":1950},\"YlxgmKTLxstUdop23VPJka\",\"collection\",\"2025-08-19T13:26:33Z\",\"backgroundColor\",{\"_23\":1955,\"_1956\":1957,\"_1958\":1959,\"_1960\":1961,\"_1970\":1971,\"_1976\":1977},\"color\",\"alpha\",1,\"hex\",\"#082c32\",\"hsl\",{\"_23\":1962,\"_1963\":1957,\"_1964\":1965,\"_1966\":1967,\"_1968\":1969},\"hslaColor\",\"a\",\"h\",188.57142857142858,\"l\",0.11372549019607843,\"s\",0.7241379310344828,\"hsv\",{\"_23\":1972,\"_1963\":1957,\"_1964\":1965,\"_1968\":1973,\"_1974\":1975},\"hsvaColor\",0.84,\"v\",0.19607843137254902,\"rgb\",{\"_23\":1978,\"_1963\":1957,\"_1979\":1980,\"_1981\":1982,\"_1983\":1984},\"rgbaColor\",\"b\",50,\"g\",44,\"r\",8,\"blackTitle\",\"AI is poised to transform how we build software at a level that echoes some of the biggest paradigm shifts in computer engineering. But as efficiencies and challenges grow, so will increased demand for solutions. \\n\\nThe future will belong to those who tackle the most important problem spaces in AI, whether we’re talking about managing compute resources, scalability, or even compliance with a still-nascent regulatory landscape.\",{\"_23\":113,\"_114\":1988},{\"_116\":1989,\"_23\":118},\"image-13c893c0d46832d4e8a48ef1a92162284c0be440-1600x1600-png\",[1991,1999,2006,2011,2017,2039],{\"_37\":1992,\"_23\":1993,\"_28\":1994,\"_1995\":1996,\"_32\":1998},\"a289f1e64649\",\"libraryItemBlock\",\"The market for AI is evolving quickly and competition grows by the day. Don't let the buzz around it trick you into thinking \\\"if we build it, they will come.\\\" But lucky for you, you don't have to worry about being the first to attempt it.\",\"item\",{\"_116\":1997,\"_23\":118},\"d2a90874-22a1-4c5f-92cd-64241af73ecd\",\"Be at the Frontier of Unproven Territory\",{\"_37\":2000,\"_23\":1993,\"_28\":2001,\"_2002\":171,\"_1995\":2003,\"_32\":2005},\"94e33378ccbe\",\"There's somewhere between 25-27 million developers worldwide and AI may add another zero, perhaps sooner than any of us think. The potentials are endless but the reality is that modern AI models aren't (yet) a silver bullet for developers.\",\"inverted\",{\"_116\":2004,\"_23\":118},\"e0c4e3c8-ab3d-4743-8cbb-d95c2b990123\",\"The Future of AI is...More Developers\",{\"_37\":2007,\"_23\":1993,\"_28\":2008,\"_1995\":2009,\"_32\":2010},\"f6d2ab1ace48\",\"With the rise of AI comes increased benefits and challenges for DevOps teams. Teams that understand how it can be leveraged to enhance current workflows and are prepared to handle its shortcomings and risks will be positioned for success.\",{\"_116\":276,\"_23\":118},\"Friend or Foe for Incident Management Teams?\",{\"_37\":2012,\"_23\":1993,\"_28\":2013,\"_2002\":171,\"_1995\":2014,\"_32\":2016},\"9e1b0ca1793f\",\"Open source and AI are both huge feats and force multipliers for developers. But when they intersect, it can get a little hairy. Stay up-to-date on the constantly shifting landscape, the different players, and what's at stake. \",{\"_116\":2015,\"_23\":118},\"8dcb096c-a91c-40b2-bcad-107add29b707\",\"“The Most Positive \u0026 Transformative Force in IT History\\\"\",{\"_37\":2018,\"_23\":2019,\"_42\":2020},\"8abb045a15fb\",\"libraryItemsBlock\",[2021,2024,2027,2030,2033,2036],{\"_37\":2022,\"_116\":2023,\"_23\":1995},\"8595e6675984\",\"6eb21538-7f94-4226-a78c-b71aa6d6412c\",{\"_37\":2025,\"_116\":2026,\"_23\":1995},\"7b8339da0d09\",\"4aff4765-dd79-4930-84e8-90d1c1c3f091\",{\"_37\":2028,\"_116\":2029,\"_23\":1995},\"f24e4a1939cb\",\"4380e6ad-2574-48f2-904d-689807ca89c7\",{\"_37\":2031,\"_116\":2032,\"_23\":1995},\"b6925450e699\",\"4af0b9f2-aa6b-4737-8bb6-12ec8d8692b6\",{\"_37\":2034,\"_116\":2035,\"_23\":1995},\"97363dce268c\",\"a700fc58-4733-408d-88c7-600e60363bc8\",{\"_37\":2037,\"_116\":2038,\"_23\":1995},\"e62be0a2ddbb\",\"07de5edb-e5bf-4de3-b6bd-4b5408fb824a\",{\"_37\":2040,\"_23\":2041,\"_2042\":2043},\"c0abbae6069d\",\"ctaModule\",\"cta\",[2044],{\"_37\":2045,\"_116\":2046,\"_23\":118},\"969649f0d150\",\"9b45536a-b953-4e63-97cc-4a3c63e43d11\",{\"_147\":2048,\"_23\":295,\"_298\":196,\"_299\":196},{\"_23\":149,\"_114\":2049},{\"_116\":2050,\"_23\":118},\"image-51b6ba1f0ae83e65e054a328e3882de99400c8b6-1600x836-png\",{\"_23\":366,\"_368\":2052},\"artificial-intelligence-for-startup-founders\",\"AI for Startup Founders\",[2055,2063,2104,2119,2159,2175,2187,2199,2207,2215,2219,2267,2275,2288,2300,2312,2320,2328,2336,2344,2352,2360,2368,2372,2387,2399,2411,2423,2431,2439,2447,2455,2463,2471,2479,2483,2498,2510,2522,2530,2538,2546,2554,2562,2570,2578,2586,2594,2598,2635,2647,2659,2671,2679,2687,2695,2703,2711,2719,2727,2735,2754,2762,2780,2788,2799,2809],{\"_37\":2056,\"_23\":313,\"_314\":2057,\"_322\":2062,\"_324\":408},\"eb11bc467dae\",[2058],{\"_37\":2059,\"_23\":318,\"_319\":2060,\"_174\":2061},\"eb3a12ad81230\",[],\"How Does Generative AI Work With Incident Response?\",[],{\"_37\":2064,\"_23\":313,\"_314\":2065,\"_322\":2097,\"_324\":325},\"eefe6d5307c6\",[2066,2070,2075,2079,2084,2088,2093],{\"_37\":2067,\"_23\":318,\"_319\":2068,\"_174\":2069},\"3369fc8096020\",[],\"Software continues to eat the world, as more dev teams depend on third-party microservices as their daily infrastructure. Which means that outages are more common and costly than ever, costing upwards of \",{\"_37\":2071,\"_23\":318,\"_319\":2072,\"_174\":2074},\"3369fc8096021\",[2073],\"390b5989751f\",\"$100K\",{\"_37\":2076,\"_23\":318,\"_319\":2077,\"_174\":2078},\"3369fc8096022\",[],\" per incident, and that successful \",{\"_37\":2080,\"_23\":318,\"_319\":2081,\"_174\":2083},\"3369fc8096023\",[2082],\"ecb6c9b7ae9b\",\"incident response workflows\",{\"_37\":2085,\"_23\":318,\"_319\":2086,\"_174\":2087},\"3369fc8096024\",[],\" are more important than ever as well. What about the wondrous wave of artificial intelligence products from Microsoft, GitHub, and OpenAI? Reports suggest generative AI tools boost \",{\"_37\":2089,\"_23\":318,\"_319\":2090,\"_174\":2092},\"3369fc8096025\",[2091],\"f7e0a6dbbe49\",\"developer productivity\",{\"_37\":2094,\"_23\":318,\"_319\":2095,\"_174\":2096},\"3369fc8096026\",[],\", reducing bottlenecks by streamlining the process of coding with code snippets, search, and summaries. Could generative AI be a breakthrough for IT operations in managing incidents that helps stem this rising tide?\",[2098,2100,2102],{\"_37\":2073,\"_23\":49,\"_439\":2099},\"https://llcbuddy.com/data/incident-management-statistics/\",{\"_37\":2082,\"_23\":49,\"_439\":2101},\"https://www.heavybit.com/devguild/incident-response\",{\"_37\":2091,\"_23\":49,\"_439\":2103},\"https://arxiv.org/abs/2302.06590\",{\"_37\":2105,\"_23\":313,\"_314\":2106,\"_322\":2116,\"_324\":325},\"a723646c9047\",[2107,2112],{\"_37\":2108,\"_23\":318,\"_319\":2109,\"_174\":2111},\"daeb999233620\",[2110],\"bbc8f76bd312\",\"62%\",{\"_37\":2113,\"_23\":318,\"_319\":2114,\"_174\":2115},\"daeb999233621\",[],\" of the general populace is “concerned” about modern AI, and 86% “believe AI could accidentally cause a catastrophic event.” So where, if at all, does GenAI fit into incident reponse and day-to-day site reliability engineering? In this article, we consulted with a panel of site incident management veterans with more than 40 years of collective experience. As one of our experts put it, GenAI is good at “confidently delivering text that is pleasant to read, but not always complete, or correct.” As another suggested, GenAI is “not good at making decisions for you...or [emulating other people’s] expertise.”\",[2117],{\"_37\":2110,\"_23\":49,\"_439\":2118},\"https://www.axios.com/2023/08/09/ai-voters-trust-government-regulation\",{\"_37\":2120,\"_23\":313,\"_314\":2121,\"_322\":2152,\"_324\":325},\"6fbe7f5e506a\",[2122,2126,2131,2134,2139,2143,2148],{\"_37\":2123,\"_23\":318,\"_319\":2124,\"_174\":2125},\"a1521bc4c9840\",[],\"Below, our panel explores known GenAI vulnerabilities in \",{\"_37\":2127,\"_23\":318,\"_319\":2128,\"_174\":2130},\"a1521bc4c9841\",[2129],\"ce8e92ee7c73\",\"security\",{\"_37\":2132,\"_23\":318,\"_319\":2133,\"_174\":846},\"a1521bc4c9842\",[],{\"_37\":2135,\"_23\":318,\"_319\":2136,\"_174\":2138},\"a1521bc4c9843\",[2137],\"759c619acc17\",\"privacy\",{\"_37\":2140,\"_23\":318,\"_319\":2141,\"_174\":2142},\"a1521bc4c9844\",[],\", not to mention its well-documented \",{\"_37\":2144,\"_23\":318,\"_319\":2145,\"_174\":2147},\"a1521bc4c9845\",[2146],\"de9b54f26917\",\"hallucinations\",{\"_37\":2149,\"_23\":318,\"_319\":2150,\"_174\":2151},\"a1521bc4c9846\",[],\", and the need for ad hoc collaboration and consequential decisions in IM. Is there an eventual future for AI-powered incident commanders, or will teams always need that proverbial human in the loop? Our panel discusses:\",[2153,2155,2157],{\"_37\":2129,\"_23\":49,\"_439\":2154},\"https://www.wired.com/story/generative-ai-prompt-injection-hacking/\",{\"_37\":2137,\"_23\":49,\"_439\":2156},\"https://iapp.org/news/a/data-protection-issues-for-employers-to-consider-when-using-generative-ai/\",{\"_37\":2146,\"_23\":49,\"_439\":2158},\"https://www.wired.com/story/fast-forward-chatbot-hallucinations-are-poisoning-web-search/\",{\"_37\":2160,\"_23\":313,\"_314\":2161,\"_2171\":1957,\"_2172\":2173,\"_322\":2174,\"_324\":325},\"4633d3d98974\",[2162,2167],{\"_37\":2163,\"_23\":318,\"_319\":2164,\"_174\":2166},\"262b6d9995b10\",[2165],\"strong\",\"The Strengths and Weaknesses of GenAI for IR and SRE:\",{\"_37\":2168,\"_23\":318,\"_319\":2169,\"_174\":2170},\"846d30174cca\",[],\" Which capabilities of GenAI are a strong fit for day-to-day incident management.\",\"level\",\"listItem\",\"bullet\",[],{\"_37\":2176,\"_23\":313,\"_314\":2177,\"_2171\":1957,\"_2172\":2173,\"_322\":2186,\"_324\":325},\"baf937bd97b7\",[2178,2182],{\"_37\":2179,\"_23\":318,\"_319\":2180,\"_174\":2181},\"964e75e826ac0\",[2165],\"How GenAI Will Affect the DevOps and SRE Professions:\",{\"_37\":2183,\"_23\":318,\"_319\":2184,\"_174\":2185},\"899855b21399\",[],\" How GenAI will impact professionals who work on both product and the operational side of product.\",[],{\"_37\":2188,\"_23\":313,\"_314\":2189,\"_2171\":1957,\"_2172\":2173,\"_322\":2198,\"_324\":325},\"840519f7d8bb\",[2190,2194],{\"_37\":2191,\"_23\":318,\"_319\":2192,\"_174\":2193},\"3a1dc2a430f90\",[2165],\"How GenAI Will Ultimately Affect Dev:\",{\"_37\":2195,\"_23\":318,\"_319\":2196,\"_174\":2197},\"ec1f1456ce11\",[],\" Our panel also weighed in with their thoughts on how GenAI will impact the general business of software development.\",[],{\"_37\":2200,\"_23\":313,\"_314\":2201,\"_322\":2206,\"_324\":325},\"a3306c00fe1c\",[2202],{\"_37\":2203,\"_23\":318,\"_319\":2204,\"_174\":2205},\"c73bcfeb284d0\",[508,2165],\"Disclaimer: While the panelists interviewed here hail from companies such as Jeli, Amazon, and PagerDuty, the views expressed below are those of the individual panelists and do not reflect the views of their employers.\",[],{\"_37\":2208,\"_23\":313,\"_314\":2209,\"_322\":2214,\"_324\":408},\"217fdd6806f9\",[2210],{\"_37\":2211,\"_23\":318,\"_319\":2212,\"_174\":2213},\"674d8be3068b0\",[],\"How to Utilize GenAI Within Incident Management Platforms\",[],{\"_37\":2216,\"_23\":149,\"_114\":2217},\"4e587cfaeee2\",{\"_116\":2218,\"_23\":118},\"image-eeaae91fe0006fb9d98aced3b6ef71737f890e98-200x200-jpg\",{\"_37\":2220,\"_23\":313,\"_314\":2221,\"_322\":2258,\"_324\":325},\"b8d85d2aec9a\",[2222,2227,2231,2236,2240,2245,2249,2254],{\"_37\":2223,\"_23\":318,\"_319\":2224,\"_174\":2226},\"8ecc46e9ad880\",[2225],\"f6f47f4e8b83\",\"Nora Jones\",{\"_37\":2228,\"_23\":318,\"_319\":2229,\"_174\":2230},\"8ecc46e9ad881\",[],\" is an incident response veteran who led IR teams at Slack and Netflix before founding the developer-first IR startup \",{\"_37\":2232,\"_23\":318,\"_319\":2233,\"_174\":2235},\"8ecc46e9ad882\",[2234],\"dddcfa115e70\",\"Jeli\",{\"_37\":2237,\"_23\":318,\"_319\":2238,\"_174\":2239},\"8ecc46e9ad883\",[],\". She’s also a co-founder of the IR community \",{\"_37\":2241,\"_23\":318,\"_319\":2242,\"_174\":2244},\"8ecc46e9ad884\",[2243],\"6d13b6e1f768\",\"LFI\",{\"_37\":2246,\"_23\":318,\"_319\":2247,\"_174\":2248},\"8ecc46e9ad885\",[],\". Her company has implemented \",{\"_37\":2250,\"_23\":318,\"_319\":2251,\"_174\":2253},\"8ecc46e9ad886\",[2252],\"22351c07f276\",\"GenAI directly into its platform\",{\"_37\":2255,\"_23\":318,\"_319\":2256,\"_174\":2257},\"8ecc46e9ad887\",[],\", utilizing natural language to rapidly spin up shareable incident reports to quickly get team members up to speed as well as to draft overarching narratives based on different touch points over an incident’s life (including detection, diagnosis, and repair moments directly from chat logs, which the platform has already been fully annotating). Jones believes that GenAI’s ability to accelerate incident logging is a valuable tool and makes GenAI worth considering as another member of the team–but not as the overarching decision maker:\",[2259,2261,2263,2265],{\"_37\":2225,\"_23\":49,\"_439\":2260},\"https://www.linkedin.com/in/norajones1/\",{\"_37\":2234,\"_23\":49,\"_439\":2262},\"https://www.jeli.io\",{\"_37\":2243,\"_23\":49,\"_439\":2264},\"https://www.learningfromincidents.io/about\",{\"_37\":2252,\"_23\":49,\"_439\":2266},\"https://techcrunch.com/2023/08/10/jeli-is-bringing-generative-ai-to-incident-report-analysis/\",{\"_37\":2268,\"_23\":313,\"_314\":2269,\"_2171\":1957,\"_2172\":2173,\"_322\":2274,\"_324\":325},\"f10746e0fce9\",[2270],{\"_37\":2271,\"_23\":318,\"_319\":2272,\"_174\":2273},\"ff5fc5ab5b310\",[2165],\"GenAI Has [At Least] Two Key Strengths for IR:\",[],{\"_37\":2276,\"_23\":313,\"_314\":2277,\"_2171\":2286,\"_2172\":2173,\"_322\":2287,\"_324\":325},\"ddb2984d384d\",[2278,2282],{\"_37\":2279,\"_23\":318,\"_319\":2280,\"_174\":2281},\"43c847de77270\",[508],\"Spinning Up Summaries to Catch Up Teams:\",{\"_37\":2283,\"_23\":318,\"_319\":2284,\"_174\":2285},\"0487ba908ddb\",[],\" As incidents are happening, GenAI’s ability to quickly spin up content can be useful to provide instantaneous summaries of incidents to relevant team members to keep everyone in the loop as things happen (rather than pulling people sideways by requiring them to drop everything and hunt down the details).\",2,[],{\"_37\":2289,\"_23\":313,\"_314\":2290,\"_2171\":2286,\"_2172\":2173,\"_322\":2299,\"_324\":325},\"dcc615e16ae7\",[2291,2295],{\"_37\":2292,\"_23\":318,\"_319\":2293,\"_174\":2294},\"6ae8b15d3bc80\",[508],\"Incident Analysis:\",{\"_37\":2296,\"_23\":318,\"_319\":2297,\"_174\":2298},\"1ab9f47e95f3\",[],\" Post-incident, GenAI can help teams uncover and compile context around incidents to create richer, more-valuable post-mortems by collecting insights and notes across the incident lifecycle.\",[],{\"_37\":2301,\"_23\":313,\"_314\":2302,\"_2171\":1957,\"_2172\":2173,\"_322\":2311,\"_324\":325},\"ca7b27531b17\",[2303,2307],{\"_37\":2304,\"_23\":318,\"_319\":2305,\"_174\":2306},\"54895b2ede500\",[2165],\"Why GenAI May Not Be Taking the Incident Commander Chair Anytime Soon:\",{\"_37\":2308,\"_23\":318,\"_319\":2309,\"_174\":2310},\"254cde6562b4\",[],\" Incident response continues to be a field full of unknowns and exceptions–not exactly a good fit for tools that are built largely to pattern-match based on past data. Fully AI-run incident remediation is unlikely to be “a thing” anytime soon.\",[],{\"_37\":2313,\"_23\":313,\"_314\":2314,\"_322\":2319,\"_324\":449},\"1c027b6bf0c8\",[2315],{\"_37\":2316,\"_23\":318,\"_319\":2317,\"_174\":2318},\"0f8b61910aa40\",[],\"Discussion: Where GenAI Makes Sense for IR with Nora Jones\",[],{\"_37\":2321,\"_23\":313,\"_314\":2322,\"_322\":2327,\"_324\":325},\"a7e405120178\",[2323],{\"_37\":2324,\"_23\":318,\"_319\":2325,\"_174\":2326},\"efeda41062f10\",[],\"In addition to offering the above observations, Jones opines that modern SRE can optimize their GenAI usage by recognizing its assorted strengths and weaknesses. Specifically, GenAI was not built on all-knowing, benevolent algorithms developed solely to decide how to manage important decisions, such as issue resolution steps. GenAI algorithms, at least for now, are optimized to generate and summarize content.\",[],{\"_37\":2329,\"_23\":313,\"_314\":2330,\"_322\":2335,\"_324\":325},\"e8b281476205\",[2331],{\"_37\":2332,\"_23\":318,\"_319\":2333,\"_174\":2334},\"b60fdb20a2ed0\",[],\"“I don't think you trust GenAI to be an ‘expert’ in anything. It's not good at making decisions for you. It's not good at [emulating other people’s] expertise. It is good at summarizing pieces of information. But just because it's easy to use and easy to ‘sprinkle’ AI on anything you're doing doesn't mean you should. I would really encourage folks that are starting to play around with it to understand actually how it works.”\",[],{\"_37\":2337,\"_23\":313,\"_314\":2338,\"_322\":2343,\"_324\":1647},\"a0b6c6257711\",[2339],{\"_37\":2340,\"_23\":318,\"_319\":2341,\"_174\":2342},\"86bdc7b348510\",[],\"I think what we really want to do is use AI to get people more curious about what's happening in their incidents.” -Nora Jones, Founder / Jeli\",[],{\"_37\":2345,\"_23\":313,\"_314\":2346,\"_322\":2351,\"_324\":325},\"8aaae63f334a\",[2347],{\"_37\":2348,\"_23\":318,\"_319\":2349,\"_174\":2350},\"47b3e090cbcd0\",[],\"“Ultimately, I think what we really want to do is use AI to get people more curious about what's happening in their incidents. I’ve always believed that if you learn how an incident actually happens, you'll be better off in the future. You can be more proactive about your incidents, resolving some of them more quickly, getting the right people in the room more quickly,” Jones explains. “Where AI seems really interesting for incident management is when we can use it to bubble up some of those interesting learnings, which then gets people investigating the incident...and gets people a little bit more curious about how it unfolded in the first place.”\",[],{\"_37\":2353,\"_23\":313,\"_314\":2354,\"_322\":2359,\"_324\":325},\"e5b372d584b2\",[2355],{\"_37\":2356,\"_23\":318,\"_319\":2357,\"_174\":2358},\"715b2d8c8e2d0\",[],\"Jones suggests that artificial intelligence provides opportunities to help both professional SREs and developers of all stripes. “I think GenAI will bring big changes in the field of incident management in terms of how incidents get communicated to stakeholders that are impacted by those incidents. But for developers in general, I think there’s an opportunity for them to use AI to accelerate their processes and help them get curious about other areas.” In the future, Jones suggests the possibility of AIs trained on large amounts of previous incident data being helpful in taking a more-proactive approach. “I don't think generative AI is going to fix the incidents for you, but I think eventually, it might help point you to previous incidents that look like the one that you're solving right now. But as far as I know, it can't get people to talk to each other. And I don’t see it being a magic box for auto-remediation anytime soon.”\",[],{\"_37\":2361,\"_23\":313,\"_314\":2362,\"_322\":2367,\"_324\":408},\"3a3e9cbcc658\",[2363],{\"_37\":2364,\"_23\":318,\"_319\":2365,\"_174\":2366},\"6f353338e7870\",[],\"Where GenAI Impacts DevOps and the Future of Software Dev\",[],{\"_37\":2369,\"_23\":149,\"_114\":2370},\"19bdb282d922\",{\"_116\":2371,\"_23\":118},\"image-c5983aa5c027d4ed93b4206b4c981083c3fcfdc4-200x200-jpg\",{\"_37\":2373,\"_23\":313,\"_314\":2374,\"_322\":2384,\"_324\":325},\"46e856c55e1e\",[2375,2380],{\"_37\":2376,\"_23\":318,\"_319\":2377,\"_174\":2379},\"1bdeba82b0320\",[2378],\"19704df303f7\",\"Jeremy Edberg\",{\"_37\":2381,\"_23\":318,\"_319\":2382,\"_174\":2383},\"1bdeba82b0321\",[],\" is a longtime DevOps expert who currently helps lead Amazon’s Alexa Operational Excellence Team, but has done tours of duty at leading tech companies including eBay, Reddit, and Netflix–where he was a founding member of Netflix’s SRE team.\",[2385],{\"_37\":2378,\"_23\":49,\"_439\":2386},\"https://www.linkedin.com/in/jedberg/\",{\"_37\":2388,\"_23\":313,\"_314\":2389,\"_2171\":1957,\"_2172\":2173,\"_322\":2398,\"_324\":325},\"1dbb3acff7c9\",[2390,2394],{\"_37\":2391,\"_23\":318,\"_319\":2392,\"_174\":2393},\"c36702b414dd0\",[2165],\"Large Language Models May Be the Next Major Evolutionary Step in Human-Computer Interfaces:\",{\"_37\":2395,\"_23\":318,\"_319\":2396,\"_174\":2397},\"e714dd32ab8f\",[],\" Whether GenAI achieves the Nirvana-like goal of artificial general intelligence (AGI), prompt-based LLM chatbots may well represent the next step in the way humans interact with technology, as they effectively help computers take a big step toward being able to understand human language.\",[],{\"_37\":2400,\"_23\":313,\"_314\":2401,\"_2171\":1957,\"_2172\":2173,\"_322\":2410,\"_324\":325},\"db8e7bfd00e0\",[2402,2406],{\"_37\":2403,\"_23\":318,\"_319\":2404,\"_174\":2405},\"cdb7c55bf0fa0\",[2165],\"We’re Not Yet at a Point Where LLMs Can Credibly Recommend Remediation Steps:\",{\"_37\":2407,\"_23\":318,\"_319\":2408,\"_174\":2409},\"cbd72b4a4f1c\",[],\" Right now, human users will trust their monitoring and alerting systems after those systems have proven themselves reliable, even going as far as allowing them to take automatic actions. But we are not there yet with LLMs. Right now, the best we can hope for is LLMs trained on previous incidents and patterns producing one (or a few) possible remediation steps and having a human select the best course of action. Over time, if the LLMs prove to produce the correct course of action in almost every case, they will be trusted to work autonomously.\",[],{\"_37\":2412,\"_23\":313,\"_314\":2413,\"_2171\":1957,\"_2172\":2173,\"_322\":2422,\"_324\":325},\"ece9d2eadad3\",[2414,2418],{\"_37\":2415,\"_23\":318,\"_319\":2416,\"_174\":2417},\"6dce022412990\",[2165],\"Potential Career Evolution for DevOps:\",{\"_37\":2419,\"_23\":318,\"_319\":2420,\"_174\":2421},\"7f5d7561f671\",[],\" Language Model Operations in AIOps?: DevOps, being generally tasked with the maintenance and caretaking of infrastructure, may also inherit the care and feeding of language models. As LLMs come to represent more-significant components in infrastructure, organizations will need people who understand distributed computing, machine learning inference, managing GPUs and CPUs next to each other, storage, and other maintenance considerations. Could there be a point where entire careers are focused on monitoring AI models, updating them, and making sure the models are getting the right inputs and appropriately learning new things?\",[],{\"_37\":2424,\"_23\":313,\"_314\":2425,\"_322\":2430,\"_324\":449},\"cd2a80857355\",[2426],{\"_37\":2427,\"_23\":318,\"_319\":2428,\"_174\":2429},\"cfe3e34a31410\",[],\"Discussion: What the Future Looks Like for Devs Using GenAI with Jeremy Edberg\",[],{\"_37\":2432,\"_23\":313,\"_314\":2433,\"_322\":2438,\"_324\":325},\"8e3e2b6a453d\",[2434],{\"_37\":2435,\"_23\":318,\"_319\":2436,\"_174\":2437},\"4df59a9902ef0\",[],\"“Right now, GenAI is something of an advisory tool. We're not to the point where we trust it enough to take the actions based on what it says,” Edberg explains. “In some ways, you could compare some of GenAI’s use cases to those of what monitoring used to be–or how things are when you're first starting out because you don't know that your monitoring and alerting are correct.”\",[],{\"_37\":2440,\"_23\":313,\"_314\":2441,\"_322\":2446,\"_324\":325},\"e00983e08adb\",[2442],{\"_37\":2443,\"_23\":318,\"_319\":2444,\"_174\":2445},\"529ff5e69a680\",[],\"“As an advisory tool, GenAI can tell you, ‘Hey, something is probably wrong here, and you should look into it,’ but there still needs to be a human in the loop there. Eventually, we'll get to the point where we can take the human out of the loop for the easy stuff...maybe. The thing is, better monitoring and alerting have already made changes to the way we operate. And LLMs will definitely make changes to the way we operate, but there'll be new challenges instead. Overall, I don’t think GenAI will eliminate DevOps jobs. It will, hopefully, make DevOps practices–and practitioners–much more efficient. So maybe in that regard, it would actually generate some net-new jobs.”\",[],{\"_37\":2448,\"_23\":313,\"_314\":2449,\"_322\":2454,\"_324\":1647},\"9c33c24fa7d1\",[2450],{\"_37\":2451,\"_23\":318,\"_319\":2452,\"_174\":2453},\"8acf5425396c\",[],\"In the future, if you are good at logic and want to learn how to reason about computer systems, [software engineering will still be] a great place to be.” -Jeremy Edberg, Principal Engineer / Amazon\",[],{\"_37\":2456,\"_23\":313,\"_314\":2457,\"_322\":2462,\"_324\":325},\"9716642e50c2\",[2458],{\"_37\":2459,\"_23\":318,\"_319\":2460,\"_174\":2461},\"d4685a3f43b90\",[],\"What effect will GenAI have on day-to-day dev workflows, or on software engineers as a profession? “If I were addressing a class of junior developers, I’d tell them, ‘GenAI is going to be a tool that will drastically speed up your development process, but it will not replace you.’ Not yet, anyway,” says Edberg. “Could it lower the barrier to entry for getting a job as an engineer? I could definitely see a situation where people–who hadn't considered this type of career before, maybe because they weren't interested in learning the details of coding syntax, for example, but are still good at general reasoning—might choose engineering now instead of business, law, or some other path.\\\"\",[],{\"_37\":2464,\"_23\":313,\"_314\":2465,\"_322\":2470,\"_324\":325},\"e87e791a1399\",[2466],{\"_37\":2467,\"_23\":318,\"_319\":2468,\"_174\":2469},\"a21774cafd06\",[],\"\\\"Somebody who has these reasoning, logic, and analytical skills might be more interested now because the ‘hard parts’ are taken care of, the syntax, the math, that kind of stuff. In the future, I think if you are good at reasoning, good at logic, and want to learn how to reason about computer systems, it's still a great place to be. If jobs do end up going away, they will be the ‘I've learned enough to know how to write code, and I'm going to spend most of my days writing basic, boilerplate' stuff,’ because the LLMs will take care of that.”\",[],{\"_37\":2472,\"_23\":313,\"_314\":2473,\"_322\":2478,\"_324\":408},\"0b261b3fa906\",[2474],{\"_37\":2475,\"_23\":318,\"_319\":2476,\"_174\":2477},\"ac4e0d817c130\",[],\"Deferring Low-Level Tasks to AI so Humans Can Focus on Strategy\",[],{\"_37\":2480,\"_23\":149,\"_114\":2481},\"02d5b29278d2\",{\"_116\":2482,\"_23\":118},\"image-c87a2249f5a3a3696fc35defa73409803afeb2a6-200x200-jpg\",{\"_37\":2484,\"_23\":313,\"_314\":2485,\"_322\":2495,\"_324\":325},\"4f595a74a5d3\",[2486,2491],{\"_37\":2487,\"_23\":318,\"_319\":2488,\"_174\":2490},\"c66b562052680\",[2489],\"dd7cc970b104\",\"Mandi Walls\",{\"_37\":2492,\"_23\":318,\"_319\":2493,\"_174\":2494},\"c66b562052681\",[],\" is a long-tenured developer advocate who has been advocating for AI and automated solutions to help make SRE and DevOps teams more productive for some time. She’s currently building communities of highly engaged developers at PagerDuty and has also served tours of duty at Chef and AOL.\",[2496],{\"_37\":2489,\"_23\":49,\"_439\":2497},\"https://www.linkedin.com/in/mandiwalls/\",{\"_37\":2499,\"_23\":313,\"_314\":2500,\"_2171\":1957,\"_2172\":2173,\"_322\":2509,\"_324\":325},\"dbdf2d9904ed\",[2501,2505],{\"_37\":2502,\"_23\":318,\"_319\":2503,\"_174\":2504},\"3508a9164e9a0\",[2165],\"We’re a Ways Off From Fully AI-Powered L1 Responders:\",{\"_37\":2506,\"_23\":318,\"_319\":2507,\"_174\":2508},\"8e4cd4bec5fa\",[],\" The sheer amount of training datasets from real, recorded incidents across a single organization required to stand up completely AI-powered Level One incident responders just doesn’t exist yet. The closest path to something similar to this in the future might be from large orgs running similar services on a very similar platform with similar runtimes, which would presumably generate incidents of a similar character with similar symptoms.\",[],{\"_37\":2511,\"_23\":313,\"_314\":2512,\"_2171\":1957,\"_2172\":2173,\"_322\":2521,\"_324\":325},\"f793b3f245cc\",[2513,2517],{\"_37\":2514,\"_23\":318,\"_319\":2515,\"_174\":2516},\"b608858530530\",[2165],\"The Most Immediate AI Opportunity in SRE May Be for Low-Level Remediation Tasks:\",{\"_37\":2518,\"_23\":318,\"_319\":2519,\"_174\":2520},\"3b15706ef23a\",[],\" Generative AI might be able to make the most immediate impact if it were trained to manage the low-level hiccups and false alarms that do not require extensive triage, and which experienced SREs resolve in minutes, anyway. There’s also strategic value in developing AI tools that can manage most low-level remediation tasks–because such tools would free up veteran SRE teams to focus more of their time and undivided attention on higher-priority incidents and post-mortems.\",[],{\"_37\":2523,\"_23\":313,\"_314\":2524,\"_322\":2529,\"_324\":449},\"9e82dab8ea81\",[2525],{\"_37\":2526,\"_23\":318,\"_319\":2527,\"_174\":2528},\"3721f0e7df3b0\",[],\"Discussion: The Division of Labor Between Human SRE and AI with Mandi Walls\",[],{\"_37\":2531,\"_23\":313,\"_314\":2532,\"_322\":2537,\"_324\":325},\"85f907947ac3\",[2533],{\"_37\":2534,\"_23\":318,\"_319\":2535,\"_174\":2536},\"f6c1d20b7fed0\",[],\"Walls suggests that the immediate value of generative AI in SRE might come from spinning up documentation and after-action reports, but also in a variety of other areas. “Our incident response process includes Zoom calls, recordings, transcripts, and Slack channels, along with charts and graphs and many other kinds of data and artifacts...it’s a slog. So there’s value in letting AI generate all the components and artifacts we need.”\",[],{\"_37\":2539,\"_23\":313,\"_314\":2540,\"_322\":2545,\"_324\":325},\"56c3d402e563\",[2541],{\"_37\":2542,\"_23\":318,\"_319\":2543,\"_174\":2544},\"bf8a22f3dea90\",[],\"Regarding how GenAI and its associated tools, such as code generators, could affect the profession of development as a whole, Walls sees opportunities in many areas for GenAI to be valuable. “Stuff like coding assistants are super interesting. Some of it is really clever and is already doing a really good job for folks doing some of that work. But as someone who uses a lot of products, I'm hoping for improved documentation—and API documentation in particular–that developers don't have to write themselves. It’d be good to see tools improve enough to automatically generate all that stuff and make it more useful.”\",[],{\"_37\":2547,\"_23\":313,\"_314\":2548,\"_322\":2553,\"_324\":325},\"bd27f196045a\",[2549],{\"_37\":2550,\"_23\":318,\"_319\":2551,\"_174\":2552},\"b292935276510\",[],\"Walls suggests that testing may be another area of opportunity for GenAI to improve development pipelines. “Another use case would be generating tests. I think there's a lot of knowledge already in that space, especially over the last 10 years as that whole practice has become more automated, and maybe it will become even more so. So maybe the work of test engineers will move more towards creating better harnesses and doing performance monitoring on the testing process rather than anything like writing a tool. Also, developers have artifact repositories. There's all this stuff that has to run together really closely. And keeping that all in line plus maintaining changes that come in from the vendors would definitely be helped by additional tooling that's a little bit smarter than what we have right now.”\",[],{\"_37\":2555,\"_23\":313,\"_314\":2556,\"_322\":2561,\"_324\":1647},\"5ccc0fa9d705\",[2557],{\"_37\":2558,\"_23\":318,\"_319\":2559,\"_174\":2560},\"cbcb5a31dc4c\",[],\"For positions like SRE that are usually more directly integrated with an engineering practice, they'll see more benefits from coding tools, which could start to learn as much about infrastructure tooling as they do about regular languages and runtime-application code.” -Mandi Walls, Developer Advocate / PagerDuty\",[],{\"_37\":2563,\"_23\":313,\"_314\":2564,\"_322\":2569,\"_324\":325},\"9817bac2933f\",[2565],{\"_37\":2566,\"_23\":318,\"_319\":2567,\"_174\":2568},\"d41f4287d5f90\",[],\"“I’m thinking about something that’s even a level up from Dependabot–which right now, will send you an email that says, ‘Hey, here's this thing that needs to be updated.’ It would be really useful to see this kind of use case broadening out to alert you that your vendor is doing an upgrade. An alert that could tell you, ‘Here's what we recommend for your specific use case.’ For example, if your cloud provider is turning off instances of your level, here's where you need to migrate...and then starting to do that work for you without having to really intervene.”\",[],{\"_37\":2571,\"_23\":313,\"_314\":2572,\"_322\":2577,\"_324\":325},\"dbe2f688d98a\",[2573],{\"_37\":2574,\"_23\":318,\"_319\":2575,\"_174\":2576},\"fcaeed8a24910\",[],\"On how GenAI may affect the business of DevOps workflows, Walls is less eager to make predictions due to variance across orgs. “DevOps jobs are different in every organization. So it's possible that GenAI tools, such as code generators, could make a difference because, in some places, a lot of those folks are writing more code. However, in other places, they're just working more in advisory positions. And then, some orgs take more of a build-and-release approach. So it's hard to stay. At the macro level, if there's going to be a deep change in what it means to work in DevOps due to generative AI, well...I'm not sure there's enough of a consensus of what a DevOps engineer should be doing to be able to say that.”\",[],{\"_37\":2579,\"_23\":313,\"_314\":2580,\"_322\":2585,\"_324\":325},\"10afd7e078b1\",[2581],{\"_37\":2582,\"_23\":318,\"_319\":2583,\"_174\":2584},\"598db06fac460\",[],\"On how GenAI will affect the business of SRE, Walls is significantly more bullish. “I think for those positions like SRE that are usually more directly integrated with an engineering practice, I think they'll see more benefits from coding tools and things like that...which potentially could start to learn as much about infrastructure tooling as they do about regular languages and runtime–application code versus infrastructure code. I’d like to see GenAI help SRE teams push forward along their golden path because so much of their infrastructure is hopefully managed as code.” And in the same way that developer technology such as containers expanded into open source with Kubernetes, there may be opportunities to see open source contribute to generative coding assistants for SREs. “I think these teams will also benefit from those same code generation tools–but they may be in Terraform or Pulumi, rather than Python/Elixir/Go/Rust.”\",[],{\"_37\":2587,\"_23\":313,\"_314\":2588,\"_322\":2593,\"_324\":408},\"46adbb984ffc\",[2589],{\"_37\":2590,\"_23\":318,\"_319\":2591,\"_174\":2592},\"cde89f849f3e0\",[],\"Why Humans in the Loop May Always Be Needed in SRE\",[],{\"_37\":2595,\"_23\":149,\"_114\":2596},\"ff61903f5fab\",{\"_116\":2597,\"_23\":118},\"image-28240d630e8fb1c07da810841e34f41fd91de686-200x200-jpg\",{\"_37\":2599,\"_23\":313,\"_314\":2600,\"_322\":2628,\"_324\":325},\"daaaabb3cab5\",[2601,2606,2610,2615,2619,2624],{\"_37\":2602,\"_23\":318,\"_319\":2603,\"_174\":2605},\"ccb0bd7bf47a0\",[2604],\"52f00b83da13\",\"Brent Chapman\",{\"_37\":2607,\"_23\":318,\"_319\":2608,\"_174\":2609},\"ccb0bd7bf47a1\",[],\" is a pioneer in what is now known as modern SRE. Throughout his career in technology, he has always also worked as a volunteer in public safety and emergency services, starting as a search-and-rescue pilot and incident commander for air search and rescue. He applied the principles he learned in emergency services to his tenure at Google, where developed the company’s internal \",{\"_37\":2611,\"_23\":318,\"_319\":2612,\"_174\":2614},\"ccb0bd7bf47a2\",[2613],\"616e5552cb59\",\"Incident Management at Google (IMAG)\",{\"_37\":2616,\"_23\":318,\"_319\":2617,\"_174\":2618},\"ccb0bd7bf47a3\",[],\" practice, and later, brought similar foundational practices to Slack. He currently runs the incident management consultancy \",{\"_37\":2620,\"_23\":318,\"_319\":2621,\"_174\":2623},\"ccb0bd7bf47a4\",[2622],\"e66bb0e4a105\",\"Great Circle Associates\",{\"_37\":2625,\"_23\":318,\"_319\":2626,\"_174\":2627},\"ccb0bd7bf47a5\",[],\".\",[2629,2631,2633],{\"_37\":2604,\"_23\":49,\"_439\":2630},\"https://www.linkedin.com/in/brentchapman/\",{\"_37\":2613,\"_23\":49,\"_439\":2632},\"https://sre.google/workbook/incident-response/\",{\"_37\":2622,\"_23\":49,\"_439\":2634},\"https://greatcircle.com/\",{\"_37\":2636,\"_23\":313,\"_314\":2637,\"_2171\":1957,\"_2172\":2173,\"_322\":2646,\"_324\":325},\"56435edba8f3\",[2638,2642],{\"_37\":2639,\"_23\":318,\"_319\":2640,\"_174\":2641},\"aeb7c676fb5d0\",[2165],\"Large Language Models Have the “Natural Language” Part Down, But May Still Lack in Other Areas:\",{\"_37\":2643,\"_23\":318,\"_319\":2644,\"_174\":2645},\"c43475645aab\",[],\" While things may certainly change in the future, LLM chatbots seem best at confidently delivering text that is pleasant to read, but not always complete, or correct, and certainly not above human verification. Today’s LLMs are sometimes wrong, but never uncertain.\",[],{\"_37\":2648,\"_23\":313,\"_314\":2649,\"_2171\":1957,\"_2172\":2173,\"_322\":2658,\"_324\":325},\"b9dfd0319569\",[2650,2654],{\"_37\":2651,\"_23\":318,\"_319\":2652,\"_174\":2653},\"40cd476dbc8f0\",[2165],\"GenAI’s Greatest Value to SRE Might Be for After Reports:\",{\"_37\":2655,\"_23\":318,\"_319\":2656,\"_174\":2657},\"cbfbd1134d86\",[],\" Post-incident phases call for a great many write-ups to document the conditions leading up to outages, the circumstances and effects of the outages, and the actions taken to resolve the outages. GenAI can certainly produce something readable and user-friendly for general audiences, though expert engineers may prefer to keep all the gory details.\",[],{\"_37\":2660,\"_23\":313,\"_314\":2661,\"_2171\":1957,\"_2172\":2173,\"_322\":2670,\"_324\":325},\"ee0377037b44\",[2662,2666],{\"_37\":2663,\"_23\":318,\"_319\":2664,\"_174\":2665},\"0fd3f4081e130\",[2165],\"Maybe There’s a Future for AI-Powered Pattern-Matching and Timeframe Planning in IM:\",{\"_37\":2667,\"_23\":318,\"_319\":2668,\"_174\":2669},\"3c72aebd49fd\",[],\" Chapman recalls his years working with highly experienced engineers who were so well-versed in their systems that they seemed to have a sixth sense when it came to browsing a series of graphs and detecting seemingly imperceptible irregularities when investigating root causes. Could AI tools eventually become “smart” enough to detect such inconsistencies? Maybe. They might be even more useful in helping size incidents and projected response time windows that correspond with severity.\",[],{\"_37\":2672,\"_23\":313,\"_314\":2673,\"_322\":2678,\"_324\":449},\"ba5d422e08c3\",[2674],{\"_37\":2675,\"_23\":318,\"_319\":2676,\"_174\":2677},\"7239f87585710\",[],\"Discussion: The Brushes May Change, but Engineers Will Still Create Art with Brent Chapman\",[],{\"_37\":2680,\"_23\":313,\"_314\":2681,\"_322\":2686,\"_324\":325},\"6931f57ba4e9\",[2682],{\"_37\":2683,\"_23\":318,\"_319\":2684,\"_174\":2685},\"c28dd3e235ad0\",[],\"As Chapman reflects on the fundamental practice of incident management, he finds few intersections with GenAI’s biggest strengths and sees many of the processes as still being fundamentally human. “In incident management, the challenge is always that we know something has gone wrong, but we don't always know what has gone wrong, or who needs to do what to fix it. The process is about figuring out the details, executing that response, getting things back to a stable situation, and then getting fully recovered to a normal state of operation. All very challenging activities that involve working under time pressure that people normally don't have, working across teams that don't routinely work together and don't know each other's capabilities and concerns and considerations and so forth.”\",[],{\"_37\":2688,\"_23\":313,\"_314\":2689,\"_322\":2694,\"_324\":325},\"1de62f39e47e\",[2690],{\"_37\":2691,\"_23\":318,\"_319\":2692,\"_174\":2693},\"96071cbd4a730\",[],\"“In our day-to-day work, we establish project teams. We spend a lot of time ‘storming and norming’ to build that whole framework of ‘how do we learn about each other and work together effectively,’ and have debates and arguments and joint planning activities that let technology companies do the amazing things they do,” Chapman offers. But in the same way that Agile methodology teams emphasize flexibility and a pragmatic approach to delivery, incident teams also need to be realistic.”\",[],{\"_37\":2696,\"_23\":313,\"_314\":2697,\"_322\":2702,\"_324\":325},\"cf9de04ffe52\",[2698],{\"_37\":2699,\"_23\":318,\"_319\":2700,\"_174\":2701},\"1b52a8d2b36b\",[],\"“Those things all take time and energy to establish. And you don't have that time and energy available during an emergency.” Downtime doesn’t just provide opportunities for collaboration–it demands collaboration. “You need to find a way to work together quickly and effectively enough for the emergency, even if it's not necessarily a great way to work together in the long run. It's very top-down, it's very authoritative, it's very hierarchical, it's very old-fashioned. But it works better in an emergency. Also, not everybody who's going to help you will be available at the same time, and certainly not at the start of the incident. You have to start responding with who's available at the time and incorporate more people over time. You need to have ways of effectively putting people to work and then putting more people in the process without disrupting the work that's already in progress, so people can come up to speed without disrupting those activities and then plug themselves in, offload, or take on some new tasks related to the emergency. And somebody has to manage and coordinate all of this and manage the communications.”\",[],{\"_37\":2704,\"_23\":313,\"_314\":2705,\"_322\":2710,\"_324\":1647},\"e48e9861485c\",[2706],{\"_37\":2707,\"_23\":318,\"_319\":2708,\"_174\":2709},\"dd6aa2fb3f2c\",[],\"[Working with GenAI is] a lot like dealing with a very junior programmer. You still have to check it. You still have to write the code. You still have to write the test.” -Brent Chapman, Principal / Great Circle Associates\",[],{\"_37\":2712,\"_23\":313,\"_314\":2713,\"_322\":2718,\"_324\":325},\"df0683fdcb84\",[2714],{\"_37\":2715,\"_23\":318,\"_319\":2716,\"_174\":2717},\"5f270c56a7810\",[],\"Chapman suggests that part of the excitement, and confusion, around AI and its benefits may come from an excessive widening or narrowing of definitions. The SRE veteran discusses automated systems that he worked on at Google for real-time ‘traffic’ (server cluster load balancing) management. “I worked on a system that tracked incident response patterns whenever there were problems with a given cluster, such that the first thing we’d do is drain incoming traffic away from that cluster, and send the traffic somewhere else that's still healthy. Realizing that this was the first thing we almost always did, we decided to automate that. But we needed some guardrails. For instance, we didn't want to drain traffic away from the last cluster in any continent, or from a cluster when there were only two clusters left. So I built this system in Python which, when it was alerted to unhealthy clusters, it would ‘think’ about draining it, run through the list of checks for that service, and decide whether or not to drain the service in that cluster. All taking place while people were still responding to their pagers. But depending on your definitions, this might be more of a style of ‘mechanical turk’-style automation rather than ‘AI,’ I suppose.”\",[],{\"_37\":2720,\"_23\":313,\"_314\":2721,\"_322\":2726,\"_324\":325},\"ff31cafe599e\",[2722],{\"_37\":2723,\"_23\":318,\"_319\":2724,\"_174\":2725},\"3e1dfbd9b0c90\",[],\"What will the future hold for software engineers? For operations, Chapman sees potential in highly-trained AI systems with the ability to take certain actions, such as taking steps to provision new systems, or adjusting configurations, autonomously. “For instance, if you tie a generative AI system to your AWS console–your control system for your cloud computing system–and you can start having a conversation with it about, let’s say, bringing online another 20% of capacity in London. And the system replies that there's not enough spare capacity in London, but it can give you 15% in London and 5% across Europe, for example. You can imagine starting to have these sorts of operative discussions with it that are going to result in things happening...under approval with your supervision, and so forth. Right now, a lot of our monitoring control systems present human operators with problems, and it's up to the operator to solve them. I think the next step is going to be to have generative AI propose solutions. Now, is it going to propose better solutions than your average new hire six months out of college? Maybe. That's going to be an interesting question.”\",[],{\"_37\":2728,\"_23\":313,\"_314\":2729,\"_322\":2734,\"_324\":325},\"205a7fc5c7d3\",[2730],{\"_37\":2731,\"_23\":318,\"_319\":2732,\"_174\":2733},\"714991d7f8ac0\",[],\"On the topic of AI’s impact on developers as a whole, Chapman still feels strongly that systems will ultimately still need humans in the loop. “I think there's going to continue to be a need for more developers. However, there's going to be a new skill set that many of them have: ‘prompting,’ basically developing prompts, asking the right questions, feeding the right data to get a useful result out of the generative AI systems they're working with. It seems comparable to art. ‘Creating art’ is going to change from knowing how to mix paints and pick a brush and applying certain physical techniques to knowing how to describe what you're looking for to the AI...so that it can generate something that looks like what you want. I'm already hearing a bunch of my programmer colleagues talking about how they’ve had good luck using ChatGPT for tasks such as providing the framework of a Python application that does such and such. It’ll write the first hundred lines of code and create the first 10 files, and basically sets up your project for you. And then you can go from there. But it's a lot like dealing with a very junior programmer who doesn't always understand your intent and sometimes just goes off into the weeds. You still have to check it. You still have to write the code. You still have to write the test. I’ve also heard some people report success generating unit tests based on inputted code (which seems kind of backward, since you’re supposed to write your tests first and write the code to make the test pass), but is that better than no unit tests at all? Yeah, probably.”\",[],{\"_37\":2736,\"_23\":313,\"_314\":2737,\"_322\":2751,\"_324\":325},\"f03b7fd250a7\",[2738,2742,2747],{\"_37\":2739,\"_23\":318,\"_319\":2740,\"_174\":2741},\"7a11aebb2d740\",[],\"“One of my favorite science fiction authors is \",{\"_37\":2743,\"_23\":318,\"_319\":2744,\"_174\":2746},\"7a11aebb2d741\",[2745],\"e7cbdbf374e7\",\"Vernor Vinge\",{\"_37\":2748,\"_23\":318,\"_319\":2749,\"_174\":2750},\"7a11aebb2d742\",[],\", who was a professor of computer science at San Diego State University. He wrote a story with a character who, for various reasons having to do with relativistic time dilation and traveling near the speed of light, ends up returning home a thousand years later, despite only aging about 10 years. He believes his skills are going to be completely out of date in this new world. And it turns out, the guy's a programmer and there is a place for his skills as basically someone who understands the system 14 layers underneath what the AIs are doing in that present day. He has an ability, by understanding those underpinnings, to bypass a lot of layers and go straight to the ‘low level,’ a bit like a programmer today who still understands assembly language. Obviously, this is kind of a simplified example, but I think there are still going to be plenty of roles for humans in technology. Someone still has to verify–to ask the questions: ‘Is this right? Is this useful? Is this complete?’ I don't see that role moving away from human judgment.”\",[2752],{\"_37\":2745,\"_23\":49,\"_439\":2753},\"https://en.wikipedia.org/wiki/Vernor_Vinge\",{\"_37\":2755,\"_23\":313,\"_314\":2756,\"_322\":2761,\"_324\":408},\"8c3185928299\",[2757],{\"_37\":2758,\"_23\":318,\"_319\":2759,\"_174\":2760},\"3ec33e0d286e0\",[],\"Conclusion\",[],{\"_37\":2763,\"_23\":313,\"_314\":2764,\"_322\":2778,\"_324\":325},\"472ba1e8fbe6\",[2765,2769,2774],{\"_37\":2766,\"_23\":318,\"_319\":2767,\"_174\":2768},\"97a64d402e6d0\",[],\"GenAI is already seeing direct applications in day-to-day SRE work. For more information and discussion on how AI may affect the future of incident management, DevOps, and the business of software development as a whole, join the \",{\"_37\":2770,\"_23\":318,\"_319\":2771,\"_174\":2773},\"97a64d402e6d1\",[2772],\"cc230a529f2d\",\"DevGuild: AI Summit\",{\"_37\":2775,\"_23\":318,\"_319\":2776,\"_174\":2777},\"97a64d402e6d2\",[],\" event.\",[2779],{\"_37\":2772,\"_23\":49,\"_439\":75},{\"_37\":2781,\"_23\":313,\"_314\":2782,\"_322\":2787,\"_324\":408},\"701eb23619a8\",[2783],{\"_37\":2784,\"_23\":318,\"_319\":2785,\"_174\":2786},\"ba3da29add990\",[],\"More Resources:\",[],{\"_37\":2789,\"_23\":313,\"_314\":2790,\"_2171\":1957,\"_2172\":2173,\"_322\":2796,\"_324\":325},\"df2f50be4a6f\",[2791],{\"_37\":2792,\"_23\":318,\"_319\":2793,\"_174\":2795},\"4d151d677cd00\",[2794],\"94c99a270c8e\",\"Article - Three Key Best Practices for Modern Incident Response\",[2797],{\"_37\":2794,\"_23\":49,\"_439\":2798},\"https://www.heavybit.com/library/article/incident-response-best-practices\",{\"_37\":2800,\"_23\":313,\"_314\":2801,\"_2171\":1957,\"_2172\":2173,\"_322\":2807,\"_324\":325},\"784420f1c8e2\",[2802],{\"_37\":2803,\"_23\":318,\"_319\":2804,\"_174\":2806},\"17656e1581d00\",[2805],\"97fc93c1e8e1\",\"On-Demand Video Series - DevGuild: Incident Response\",[2808],{\"_37\":2805,\"_23\":49,\"_439\":2101},{\"_37\":2810,\"_23\":313,\"_314\":2811,\"_2171\":1957,\"_2172\":2173,\"_322\":2817,\"_324\":325},\"eb0c700a9de0\",[2812],{\"_37\":2813,\"_23\":318,\"_319\":2814,\"_174\":2816},\"ec5a9ee2c8360\",[2815],\"5bc485daa9db\",\"Podcast - Getting There with Nora Jones, Niall Murphy and Laura De Vesine\",[2818],{\"_37\":2815,\"_23\":49,\"_439\":2819},\"https://www.heavybit.com/library/podcasts/getting-there/ep-7-the-march-2023-datadog-outage-with-laura-de-vesine\",\"duration\",31,[2823],{\"_37\":2824,\"_23\":313,\"_314\":2825,\"_322\":2830,\"_324\":325},\"f1f537ffdefa\",[2826],{\"_37\":2827,\"_23\":318,\"_319\":2828,\"_174\":2829},\"1a565f1c58e50\",[],\"DevOps and IM experts from Jeli, PagerDuty, AWS, and other leading outfits explain how generative AI will affect the future of site reliability engineering.\",[],\"library\",{\"_42\":2833,\"_4116\":4117},[2834,3622,3684],{\"_398\":2835,\"_2820\":3615,\"_149\":3616,\"_3617\":3618,\"_912\":3619,\"_366\":3620,\"_32\":2963,\"_3621\":278},[2836,2844,2893,2901,2917,2925,2949,2957,2965,2981,2997,3005,3013,3058,3066,3070,3078,3086,3094,3110,3118,3126,3170,3179,3187,3195,3214,3222,3241,3249,3257,3265,3273,3281,3300,3308,3316,3335,3343,3351,3359,3409,3417,3425,3433,3441,3449,3465,3501,3509,3517,3535,3543,3551,3562,3573,3584,3594,3604],{\"_37\":2837,\"_23\":313,\"_314\":2838,\"_322\":2843,\"_324\":408},\"0ed3b58e993b\",[2839],{\"_37\":2840,\"_23\":318,\"_319\":2841,\"_174\":2842},\"4addb2e5d7580\",[],\"Why Data Pipelines and Inference Are AI Infrastructure’s Biggest Challenges\",[],{\"_37\":2845,\"_23\":313,\"_314\":2846,\"_322\":2885,\"_324\":325},\"91a2a347584c\",[2847,2851,2856,2860,2865,2868,2873,2877,2882],{\"_37\":2848,\"_23\":318,\"_319\":2849,\"_174\":2850},\"1e814d1f8b1d0\",[],\"While there’s still great excitement around AI and machine learning (ML), we’re starting to see some level of organizational maturity as enterprises work to stand up programs internally. About \",{\"_37\":2852,\"_23\":318,\"_319\":2853,\"_174\":2855},\"1e814d1f8b1d1\",[2854],\"c460fb2ba20f\",\"40% of enterprises surveyed\",{\"_37\":2857,\"_23\":318,\"_319\":2858,\"_174\":2859},\"1e814d1f8b1d2\",[],\" report they have either actively deployed an AI program or are currently exploring one. This means enterprises are now grappling with real-world challenges like \",{\"_37\":2861,\"_23\":318,\"_319\":2862,\"_174\":2864},\"1e814d1f8b1d3\",[2863],\"b81f4641f3bd\",\"building products on top of AI\",{\"_37\":2866,\"_23\":318,\"_319\":2867,\"_174\":679},\"1e814d1f8b1d4\",[],{\"_37\":2869,\"_23\":318,\"_319\":2870,\"_174\":2872},\"1e814d1f8b1d5\",[2871],\"d41ea8546d9e\",\"aligning disparate teams around their programs\",{\"_37\":2874,\"_23\":318,\"_319\":2875,\"_174\":2876},\"1e814d1f8b1d6\",[],\", and even \",{\"_37\":2878,\"_23\":318,\"_319\":2879,\"_174\":2881},\"1e814d1f8b1d7\",[2880],\"2ff30b826f86\",\"going to market\",{\"_37\":2883,\"_23\":318,\"_319\":2884,\"_174\":2627},\"1e814d1f8b1d8\",[],[2886,2888,2890,2891],{\"_37\":2854,\"_23\":49,\"_439\":2887},\"https://newsroom.ibm.com/2024-01-10-Data-Suggests-Growth-in-Enterprise-Adoption-of-AI-is-Due-to-Widespread-Deployment-by-Early-Adopters#:~:text=%2D%20About%2042%25%20of%20enterprise%2D,have%20not%20deployed%20their%20models.\",{\"_37\":2863,\"_23\":49,\"_439\":2889},\"https://www.heavybit.com/library/article/prompting-chatgpt-programming-llm-software-development\",{\"_37\":2871,\"_23\":49,\"_439\":631},{\"_37\":2880,\"_23\":49,\"_439\":2892},\"https://www.heavybit.com/library/video/from-labs-to-launch-stories-from-github-copilot\",{\"_37\":2894,\"_23\":313,\"_314\":2895,\"_322\":2900,\"_324\":325},\"61d2c20b2b35\",[2896],{\"_37\":2897,\"_23\":318,\"_319\":2898,\"_174\":2899},\"2906b04616c70\",[],\"More pointedly, foundational challenges in building and running the necessary infrastructure to underpin such programs are beginning to emerge. The ideal future for enterprises is having the ability to run successful programs powered by a large language model (LLM) that provide significant competitive advantage compared to programs powered by off-the-shelf, open-weight models utilizing the white-labeled OEM dataset provided with purchase. Generally speaking, typing a generic prompt into a generic model will only ever produce a generic, low-value response.\",[],{\"_37\":2902,\"_23\":313,\"_314\":2903,\"_322\":2916,\"_324\":325},\"743619ad3094\",[2904,2908,2912],{\"_37\":2905,\"_23\":318,\"_319\":2906,\"_174\":2907},\"671a92cd342a0\",[],\"Instead, the ideal future for a best-in-class program involves having the appropriate model (or \",{\"_37\":2909,\"_23\":318,\"_319\":2910,\"_174\":2911},\"671a92cd342a1\",[508],\"models\",{\"_37\":2913,\"_23\":318,\"_319\":2914,\"_174\":2915},\"671a92cd342a2\",[],\") humming away in production, running in an affordable, easy-to-manage configuration, and continually running jobs using a pristine, secure, first-party dataset. But getting to our perfect picture is much easier said than done.\",[],{\"_37\":2918,\"_23\":313,\"_314\":2919,\"_322\":2924,\"_324\":325},\"f4b781a68f3b\",[2920],{\"_37\":2921,\"_23\":318,\"_319\":2922,\"_174\":2923},\"71aee732ef660\",[],\"As far as I can tell, this is the core of the modern infrastructure problem for enterprises:\",[],{\"_37\":2926,\"_23\":313,\"_314\":2927,\"_322\":2948,\"_324\":1647},\"8bc666e9c481\",[2928,2932,2936,2940,2944],{\"_37\":2929,\"_23\":318,\"_319\":2930,\"_174\":2931},\"bd2fc30ade430\",[508],\"The biggest challenge emerging is building and operating the infrastructure both for creating and running the \",{\"_37\":2933,\"_23\":318,\"_319\":2934,\"_174\":2935},\"bd2fc30ade431\",[],\"data pipelines\",{\"_37\":2937,\"_23\":318,\"_319\":2938,\"_174\":2939},\"bd2fc30ade432\",[508],\" to build, manage, and maintain a robust, secure body of proprietary data to train, fine-tune, and orchestrate LLM operations, and for running \",{\"_37\":2941,\"_23\":318,\"_319\":2942,\"_174\":2943},\"bd2fc30ade433\",[],\"inference,\",{\"_37\":2945,\"_23\":318,\"_319\":2946,\"_174\":2947},\"bd2fc30ade434\",[508],\" the actual process of models running calculations on inputted data.\",[],{\"_37\":2950,\"_23\":313,\"_314\":2951,\"_322\":2956,\"_324\":325},\"a2dd92e48448\",[2952],{\"_37\":2953,\"_23\":318,\"_319\":2954,\"_174\":2955},\"411316f226290\",[],\"Let’s unpack the situation.\",[],{\"_37\":2958,\"_23\":313,\"_314\":2959,\"_322\":2964,\"_324\":408},\"9ecae1e4aa02\",[2960],{\"_37\":2961,\"_23\":318,\"_319\":2962,\"_174\":2963},\"e753ca1c205f0\",[],\"The Data Pipeline is the New Secret Sauce\",[],{\"_37\":2966,\"_23\":313,\"_314\":2967,\"_322\":2980,\"_324\":325},\"7d5330589a09\",[2968,2972,2976],{\"_37\":2969,\"_23\":318,\"_319\":2970,\"_174\":2971},\"434f33f31ef50\",[],\"The enterprise data pipeline is unique and different from what startups are able to do because of the organizational, regulatory, and customer challenges they face. And so, data pipelines that \",{\"_37\":2973,\"_23\":318,\"_319\":2974,\"_174\":2975},\"434f33f31ef51\",[508],\"work\",{\"_37\":2977,\"_23\":318,\"_319\":2978,\"_174\":2979},\"434f33f31ef52\",[],\" represent a competitive advantage for their organizations. And they also, by necessity, are slow to evolve because the risks of those very same things are commensurate with their value. So, the similarities between this situation and the emergence of DevOps are 1:1–these are precisely the conditions that emerged when I was at Amazon and observed other organizations trying to survive and scale.\",[],{\"_37\":2982,\"_23\":313,\"_314\":2983,\"_322\":2996,\"_324\":325},\"792c03d45604\",[2984,2988,2992],{\"_37\":2985,\"_23\":318,\"_319\":2986,\"_174\":2987},\"624177888d260\",[],\"What I think is different–and this is the important part–is that unlike a lot of the early DevOps concepts, everyone was on a journey to CI/CD. Whereas today, we are \",{\"_37\":2989,\"_23\":318,\"_319\":2990,\"_174\":2991},\"624177888d261\",[508],\"starting from\",{\"_37\":2993,\"_23\":318,\"_319\":2994,\"_174\":2995},\"624177888d262\",[],\" this point with the data pipeline.\",[],{\"_37\":2998,\"_23\":313,\"_314\":2999,\"_322\":3004,\"_324\":449},\"b8a46f1e3762\",[3000],{\"_37\":3001,\"_23\":318,\"_319\":3002,\"_174\":3003},\"60d10311d9aa0\",[],\"The Data “DevOps Moment”\",[],{\"_37\":3006,\"_23\":313,\"_314\":3007,\"_322\":3012,\"_324\":325},\"7a81e6f0695f\",[3008],{\"_37\":3009,\"_23\":318,\"_319\":3010,\"_174\":3011},\"99421d91e0db0\",[],\"The ability to continuously create and improve high-quality, first-party training datasets and then develop and finetune models are a strategic advantage.\",[],{\"_37\":3014,\"_23\":313,\"_314\":3015,\"_322\":3051,\"_324\":325},\"46a8a75c6b93\",[3016,3020,3025,3029,3034,3038,3042,3047],{\"_37\":3017,\"_23\":318,\"_319\":3018,\"_174\":3019},\"d2ecbdabaf9a0\",[],\"At \",{\"_37\":3021,\"_23\":318,\"_319\":3022,\"_174\":3024},\"d2ecbdabaf9a1\",[3023],\"1c7659c99b97\",\"Data Council\",{\"_37\":3026,\"_23\":318,\"_319\":3027,\"_174\":3028},\"d2ecbdabaf9a2\",[],\" this year, I said that data pipelines are having a “DevOps” moment, starting with a cultural and technical shift toward \",{\"_37\":3030,\"_23\":318,\"_319\":3031,\"_174\":3033},\"d2ecbdabaf9a3\",[3032,508],\"b27515ec2a3b\",\"continuous integration/continuous delivery\",{\"_37\":3035,\"_23\":318,\"_319\":3036,\"_174\":3037},\"d2ecbdabaf9a4\",[508],\" \",{\"_37\":3039,\"_23\":318,\"_319\":3040,\"_174\":3041},\"d2ecbdabaf9a5\",[],\"(CI/CD). This is now accelerating as AI programs in production are running data jobs through their models, resulting in outputs of increasing quality and power. (I also said that the only reason this hadn’t happened before is that data people seem to be more introverted than we were at the beginning of the DevOps moment.) Creating and maintaining systems that are this complex requires a dedicated team and resources–Gartner reports that \",{\"_37\":3043,\"_23\":318,\"_319\":3044,\"_174\":3046},\"d2ecbdabaf9a6\",[3045],\"937945d249d3\",\"87%\",{\"_37\":3048,\"_23\":318,\"_319\":3049,\"_174\":3050},\"d2ecbdabaf9a7\",[],\" of “mature organizations” have dedicated AI teams. These types of projects aren’t accidental. They cannot be bought off the shelf. Building out a functional data pipeline for AI programs will be a valuable advantage now, and will become more so every day. Each enterprise’s internal dataset is an artifact–the end result of a series of important data management processes run through a complicated data toolchain. Creating this data pipeline is itself a massive barrier that requires significant operational effectiveness combined with security and privacy practices that prevent any personally identifiable information (PII) from “leaking.”\",[3052,3054,3056],{\"_37\":3023,\"_23\":49,\"_439\":3053},\"https://x.com/jesserobbins/status/1772699474827034692/photo/1\",{\"_37\":3032,\"_23\":49,\"_439\":3055},\"https://en.wikipedia.org/wiki/CI/CD\",{\"_37\":3045,\"_23\":49,\"_439\":3057},\"https://www.gartner.com/en/newsroom/press-releases/2024-05-14-artificial-intelligence-is-creating-new-roles-and-skills-in-data-and-analytics\",{\"_37\":3059,\"_23\":313,\"_314\":3060,\"_322\":3065,\"_324\":325},\"64628336d51b\",[3061],{\"_37\":3062,\"_23\":318,\"_319\":3063,\"_174\":3064},\"708f669d0f230\",[],\"At best, without an appropriate data pipeline, enterprises will simply fail to build their ideal internal dataset. At worst, they incur significant business risk from a variety of data-related factors, including privacy leaks, subpar performance that produces inaccurate results, and the potentially high costs of having to re-train models initially trained on poor-quality data (to name just a few).\",[],{\"_37\":3067,\"_23\":149,\"_114\":3068},\"58e308103974\",{\"_116\":3069,\"_23\":118},\"image-68199dbf4962726476f88bfb0a1454f2318e1035-1600x808-jpg\",{\"_37\":3071,\"_23\":313,\"_314\":3072,\"_322\":3077,\"_324\":325},\"7952ee5b5bad\",[3073],{\"_37\":3074,\"_23\":318,\"_319\":3075,\"_174\":3076},\"d5122389e9d70\",[508],\"Data management for AI is having its DevOps moment.\",[],{\"_37\":3079,\"_23\":313,\"_314\":3080,\"_322\":3085,\"_324\":1647},\"438cc2914cf6\",[3081],{\"_37\":3082,\"_23\":318,\"_319\":3083,\"_174\":3084},\"ed0927bdb8120\",[],\"The data pipeline for internal AI programs will need to be a continuous process that doesn’t “end” when you launch your LLM into production, but rather, begins at that point–and requires continuous iteration, refinement, and monitoring, like any other software product. In order to master the process of managing AI data–which gives them the ability to adapt to change and drive higher performance over time–enterprises will need to invest in cultivating team-wide, organizational capability to both build and maintain their data pipeline.\",[],{\"_37\":3087,\"_23\":313,\"_314\":3088,\"_322\":3093,\"_324\":325},\"536a1c6c67a1\",[3089],{\"_37\":3090,\"_23\":318,\"_319\":3091,\"_174\":3092},\"233e17cfca220\",[],\"To clarify, much has been written about MLOps tooling and the exciting shifts happening there–so while we won’t be doing a deep dive here, I do recommend looking into the space if you’re interested.\",[],{\"_37\":3095,\"_23\":313,\"_314\":3096,\"_322\":3109,\"_324\":325},\"6b98d17ad841\",[3097,3101,3105],{\"_37\":3098,\"_23\":318,\"_319\":3099,\"_174\":3100},\"fcc11a63ad310\",[],\"However, in addition to creating their data pipeline, organizations are also running into another challenge that we’ve seen over and over again: Once you \",{\"_37\":3102,\"_23\":318,\"_319\":3103,\"_174\":3104},\"fcc11a63ad311\",[508],\"have\",{\"_37\":3106,\"_23\":318,\"_319\":3107,\"_174\":3108},\"fcc11a63ad312\",[],\" built a data pipeline, what do you do with it?\",[],{\"_37\":3111,\"_23\":313,\"_314\":3112,\"_322\":3117,\"_324\":325},\"f640efb67018\",[3113],{\"_37\":3114,\"_23\":318,\"_319\":3115,\"_174\":3116},\"4af7c8236de40\",[],\"You need to be able to run inference on your data processes securely and privately, at scale–and without breaking the bank.\",[],{\"_37\":3119,\"_23\":313,\"_314\":3120,\"_322\":3125,\"_324\":408},\"cc244f004f89\",[3121],{\"_37\":3122,\"_23\":318,\"_319\":3123,\"_174\":3124},\"fd64076b11540\",[],\"The Inference Hosting Challenge\",[],{\"_37\":3127,\"_23\":313,\"_314\":3128,\"_322\":3163,\"_324\":325},\"4fa8538162a9\",[3129,3133,3138,3141,3145,3150,3154,3159],{\"_37\":3130,\"_23\":318,\"_319\":3131,\"_174\":3132},\"6f88378a00bc0\",[],\"Inference for LLMs requires computing resources, most commonly provided by high-powered \",{\"_37\":3134,\"_23\":318,\"_319\":3135,\"_174\":3137},\"6f88378a00bc1\",[3136,508],\"e3d0ba71f902\",\"AI accelerator\",{\"_37\":3139,\"_23\":318,\"_319\":3140,\"_174\":3037},\"6f88378a00bc2\",[508],{\"_37\":3142,\"_23\":318,\"_319\":3143,\"_174\":3144},\"6f88378a00bc3\",[],\"hardware most commonly found in high-end \",{\"_37\":3146,\"_23\":318,\"_319\":3147,\"_174\":3149},\"6f88378a00bc4\",[3148,508],\"0bace830c933\",\"GPU chips\",{\"_37\":3151,\"_23\":318,\"_319\":3152,\"_174\":3153},\"6f88378a00bc5\",[],\". Hardware that is powerful enough to run inference at scale is, for the time being, both \",{\"_37\":3155,\"_23\":318,\"_319\":3156,\"_174\":3158},\"6f88378a00bc6\",[3157],\"297061c00560\",\"costly and rare\",{\"_37\":3160,\"_23\":318,\"_319\":3161,\"_174\":3162},\"6f88378a00bc7\",[],\". Due to the variety of challenges, we are seeing organizations adopting one of the following inference hosting models:\",[3164,3166,3168],{\"_37\":3136,\"_23\":49,\"_439\":3165},\"https://en.wikipedia.org/wiki/AI_accelerator\",{\"_37\":3148,\"_23\":49,\"_439\":3167},\"https://en.wikipedia.org/wiki/Graphics_processing_unit\",{\"_37\":3157,\"_23\":49,\"_439\":3169},\"https://www.cnn.com/2023/08/06/tech/ai-chips-supply-chain/index.html\",{\"_37\":3171,\"_23\":313,\"_314\":3172,\"_322\":3177,\"_324\":3178},\"4a6057760be1\",[3173],{\"_37\":3174,\"_23\":318,\"_319\":3175,\"_174\":3176},\"dcce28d70b2d0\",[508],\"1. Hosted Inference via API\",[],\"h4\",{\"_37\":3180,\"_23\":313,\"_314\":3181,\"_322\":3186,\"_324\":325},\"c6f7eda75d1c\",[3182],{\"_37\":3183,\"_23\":318,\"_319\":3184,\"_174\":3185},\"fdc74d6d66bf0\",[],\"Many enterprises are choosing to work with third-party API providers such as OpenAI and Anthropic. External providers host LLM models and inference burdens themselves–abstracting away the complexity and cost into token allowances across various pricing tiers, without the need to invest significant capital expenditure into acquiring AI accelerator hardware, or the need to invest operating expenditure into hiring a full-time team on call to run a data center.\",[],{\"_37\":3188,\"_23\":313,\"_314\":3189,\"_322\":3194,\"_324\":3178},\"61525f070001\",[3190],{\"_37\":3191,\"_23\":318,\"_319\":3192,\"_174\":3193},\"33323dc445dc0\",[508],\"2. On-Device “Edge” Hosting\",[],{\"_37\":3196,\"_23\":313,\"_314\":3197,\"_322\":3211,\"_324\":325},\"e8cd6b11e69d\",[3198,3202,3207],{\"_37\":3199,\"_23\":318,\"_319\":3200,\"_174\":3201},\"3108e08e22120\",[],\"An emerging configuration for smaller teams is \",{\"_37\":3203,\"_23\":318,\"_319\":3204,\"_174\":3206},\"3108e08e22121\",[3205,508],\"bb3193d1684a\",\"edge computing\",{\"_37\":3208,\"_23\":318,\"_319\":3209,\"_174\":3210},\"3108e08e22122\",[],\"–specifically, bringing LLMs “closer” to data sources by hosting AI models locally on-device. By pairing “smaller” models (in the range of ~3 billion parameters) with high-end laptop computers, organizations can see reduced latency, better bandwidth usage, and improved privacy by keeping data local–though it’s not clear this configuration can scale well for larger teams that may need larger models.\",[3212],{\"_37\":3205,\"_23\":49,\"_439\":3213},\"https://en.wikipedia.org/wiki/Edge_computing\",{\"_37\":3215,\"_23\":313,\"_314\":3216,\"_322\":3221,\"_324\":3178},\"4705a879c633\",[3217],{\"_37\":3218,\"_23\":318,\"_319\":3219,\"_174\":3220},\"55d7973350520\",[508],\"3. On-Premise Data Center\",[],{\"_37\":3223,\"_23\":313,\"_314\":3224,\"_322\":3238,\"_324\":325},\"6791eea9b508\",[3225,3229,3234],{\"_37\":3226,\"_23\":318,\"_319\":3227,\"_174\":3228},\"75366fb120d80\",[],\"There was a time when running an on-premise data center was so crucial to operations that many enterprises owned and ran their own. However, as of \",{\"_37\":3230,\"_23\":318,\"_319\":3231,\"_174\":3233},\"75366fb120d81\",[3232],\"f574da1b2098\",\"2023\",{\"_37\":3235,\"_23\":318,\"_319\":3236,\"_174\":3237},\"75366fb120d82\",[],\", the majority of enterprise IT workloads are now hosted off-premise due to the massive total cost of ownership, including hardware, physical real estate, and operational teams on call. It’s likely that enterprises will continue to prefer externally-hosted solutions for inference as well.\",[3239],{\"_37\":3232,\"_23\":49,\"_439\":3240},\"https://journal.uptimeinstitute.com/the-majority-of-enterprise-it-is-now-off-premises/\",{\"_37\":3242,\"_23\":313,\"_314\":3243,\"_322\":3248,\"_324\":3178},\"2ecd41f07377\",[3244],{\"_37\":3245,\"_23\":318,\"_319\":3246,\"_174\":3247},\"196d985177470\",[508],\"4. Off-Premise Cloud Hosting via Third-Party Data Center\",[],{\"_37\":3250,\"_23\":313,\"_314\":3251,\"_322\":3256,\"_324\":325},\"290f85725673\",[3252],{\"_37\":3253,\"_23\":318,\"_319\":3254,\"_174\":3255},\"3d2ce89b21900\",[],\"Third-party data center hosting for AI inference is increasingly beginning to resemble hosting for traditional cloud computing, and it’s possible that in the future, we’ll see a core group of leading vendors in third-party AI inference hosting that will gain popularity among enterprises. However, dependence on externally-hosted resources could also introduce latency into larger-scale inference jobs, dependencies on external data center performance, and external security and privacy threats.\",[],{\"_37\":3258,\"_23\":313,\"_314\":3259,\"_322\":3264,\"_324\":408},\"5fefa5c2ce98\",[3260],{\"_37\":3261,\"_23\":318,\"_319\":3262,\"_174\":3263},\"e24e130278070\",[],\"Getting Enterprise Ready for AI\",[],{\"_37\":3266,\"_23\":313,\"_314\":3267,\"_322\":3272,\"_324\":325},\"db06cb5731eb\",[3268],{\"_37\":3269,\"_23\":318,\"_319\":3270,\"_174\":3271},\"32dafb13b9cc0\",[],\"Given the challenges and costs associated with launching an enterprise-scale AI program, we’re starting to see discrete phases emerge:\",[],{\"_37\":3274,\"_23\":313,\"_314\":3275,\"_322\":3280,\"_324\":3178},\"91ff98ffe685\",[3276],{\"_37\":3277,\"_23\":318,\"_319\":3278,\"_174\":3279},\"bd8104e5b74d0\",[508],\"Phase 1: Starting a Program With an off-the-shelf Cloud Provider\",[],{\"_37\":3282,\"_23\":313,\"_314\":3283,\"_322\":3297,\"_324\":325},\"70d8c3ead43a\",[3284,3288,3293],{\"_37\":3285,\"_23\":318,\"_319\":3286,\"_174\":3287},\"f862b42693180\",[],\"Unless an enterprise is restricted to on-prem and has a mature infrastructure capability already, most organizations are going to start serious experimentation with existing Cloud provider(s) using currently available APIs. For example, Microsoft has reported \",{\"_37\":3289,\"_23\":318,\"_319\":3290,\"_174\":3292},\"f862b42693181\",[3291],\"58c2bf665d06\",\"53,000\",{\"_37\":3294,\"_23\":318,\"_319\":3295,\"_174\":3296},\"f862b42693182\",[],\" organizations using its AI offerings via Azure.\",[3298],{\"_37\":3291,\"_23\":49,\"_439\":3299},\"https://azure.microsoft.com/en-us/blog/celebrating-customers-journeys-to-ai-innovation-at-microsoft-build-2024/\",{\"_37\":3301,\"_23\":313,\"_314\":3302,\"_322\":3307,\"_324\":325},\"4e1b88654b99\",[3303],{\"_37\":3304,\"_23\":318,\"_319\":3305,\"_174\":3306},\"19f0060f6be90\",[],\"At this stage, data science and operations teams will focus on testing and implementing valuable use cases and putting them into production via their providers’ models. Inference burdens are abstracted away as part of existing contracts with established providers. Smart enterprises will of course do everything they can to maintain as high a level of privacy and security as possible around the data they feed in. However, they may face headwinds due to possible reverberations around data leakage–there are many counternarratives, and fear around, having third-party models trained on proprietary data.\",[],{\"_37\":3309,\"_23\":313,\"_314\":3310,\"_322\":3315,\"_324\":3178},\"65b8ed1d041a\",[3311],{\"_37\":3312,\"_23\":318,\"_319\":3313,\"_174\":3314},\"bfe2c7419f950\",[508],\"Phase 2: Scaling the Existing Solution\",[],{\"_37\":3317,\"_23\":313,\"_314\":3318,\"_322\":3332,\"_324\":325},\"b98dc8857710\",[3319,3323,3328],{\"_37\":3320,\"_23\":318,\"_319\":3321,\"_174\":3322},\"2c1e6a87d18b\",[],\"Over time, enterprises will stand up a functional data pipeline, which may not be perfect, but will be performant enough to zero in on valuable use cases that deliver outsize value (like \",{\"_37\":3324,\"_23\":318,\"_319\":3325,\"_174\":3327},\"bfe2c7419f952\",[3326],\"3a42c19f340e\",\"batch processing jobs\",{\"_37\":3329,\"_23\":318,\"_319\":3330,\"_174\":3331},\"bfe2c7419f953\",[],\" for large quantities of data). As they do, they may feel compelled to improve their data privacy posture, at least for certain types of data and certain data processing jobs.\",[3333],{\"_37\":3326,\"_23\":49,\"_439\":3334},\"https://www.oreilly.com/library/view/artificial-intelligence-for/9781788472173/9f419a4f-27bc-48ca-afbc-71006e79f6cb.xhtml\",{\"_37\":3336,\"_23\":313,\"_314\":3337,\"_322\":3342,\"_324\":325},\"c4e96f137a38\",[3338],{\"_37\":3339,\"_23\":318,\"_319\":3340,\"_174\":3341},\"0ac4a73b26220\",[],\"Also, as enterprises begin to realize value from their developing AI program, they will also assess the growing costs of their API provider contracts and tooling. At this stage, enterprises will look to make significant optimizations on their usage and, in the interest of reducing costs, they focus on efficiency.\",[],{\"_37\":3344,\"_23\":313,\"_314\":3345,\"_322\":3350,\"_324\":3178},\"7e19b4489c2c\",[3346],{\"_37\":3347,\"_23\":318,\"_319\":3348,\"_174\":3349},\"5b081dd2f9c30\",[508],\"Phase 3: Cost Shock and Optimization\",[],{\"_37\":3352,\"_23\":313,\"_314\":3353,\"_322\":3358,\"_324\":325},\"b1338ba270e2\",[3354],{\"_37\":3355,\"_23\":318,\"_319\":3356,\"_174\":3357},\"5f44c02a61eb0\",[],\"Specifically, enterprises may find themselves weighing the cost and complexity of standing up an internal LLM against the benefits of maintaining a fully closed-circuit dataset they no longer have to feed to an externally-hosted system–freeing themselves from external model risks such as outages or security breaches experienced by their vendor, and from massive provider bills. As they grow their capabilities in managing a ML program production, they will also likely grow their knowledge of hosting models internally and will seriously consider hosting and launching their own internal model, powered by an inference configuration that makes operational and financial sense.\",[],{\"_37\":3360,\"_23\":313,\"_314\":3361,\"_322\":3400,\"_324\":325},\"f81a38f97e23\",[3362,3366,3371,3374,3379,3382,3387,3391,3396],{\"_37\":3363,\"_23\":318,\"_319\":3364,\"_174\":3365},\"f41c5344a7350\",[],\"At such a point, enterprises may start thinking about investing in tooling such as \",{\"_37\":3367,\"_23\":318,\"_319\":3368,\"_174\":3370},\"f41c5344a7351\",[3369,508],\"78437f7bf1f9\",\"pretraining datasets\",{\"_37\":3372,\"_23\":318,\"_319\":3373,\"_174\":679},\"f41c5344a7352\",[],{\"_37\":3375,\"_23\":318,\"_319\":3376,\"_174\":3378},\"f41c5344a7353\",[3377,508],\"10188bf6a62a\",\"data filtering\",{\"_37\":3380,\"_23\":318,\"_319\":3381,\"_174\":846},\"f41c5344a7354\",[],{\"_37\":3383,\"_23\":318,\"_319\":3384,\"_174\":3386},\"f41c5344a7355\",[3385,508],\"a638c087e8a5\",\"splitting\",{\"_37\":3388,\"_23\":318,\"_319\":3389,\"_174\":3390},\"f41c5344a7356\",[],\", and \",{\"_37\":3392,\"_23\":318,\"_319\":3393,\"_174\":3395},\"f41c5344a7357\",[3394,508],\"a319ff2cf553\",\"model evaluation\",{\"_37\":3397,\"_23\":318,\"_319\":3398,\"_174\":3399},\"f41c5344a7358\",[],\" to select a model that makes sense for them. They’ll also need to seriously consider inference hosting configurations, potentially opting for edge computing with a small test team to begin with and migrating to an externally-hosted cloud inference provider as their needs increase in scale.\",[3401,3403,3405,3407],{\"_37\":3369,\"_23\":49,\"_439\":3402},\"https://en.wikipedia.org/wiki/Training,_validation,_and_test_data_sets#:~:text=A%20training%20data%20set%20is,%2C%20for%20example%2C%20a%20classifier.\",{\"_37\":3377,\"_23\":49,\"_439\":3404},\"https://machinelearning.apple.com/research/data-filtering-networks\",{\"_37\":3385,\"_23\":49,\"_439\":3406},\"https://www.techtarget.com/searchenterpriseai/definition/data-splitting\",{\"_37\":3394,\"_23\":49,\"_439\":3408},\"https://towardsdatascience.com/various-ways-to-evaluate-a-machine-learning-models-performance-230449055f15?gi=aa83f96d4174\",{\"_37\":3410,\"_23\":313,\"_314\":3411,\"_322\":3416,\"_324\":325},\"924da34e6f79\",[3412],{\"_37\":3413,\"_23\":318,\"_319\":3414,\"_174\":3415},\"a97fb46564930\",[],\"As their familiarity with running AI operations internally grows, they also begin to more-strongly eye the value of closed-circuit privacy offered by running their own models internally–without the help of an off-the-shelf provider.\",[],{\"_37\":3418,\"_23\":313,\"_314\":3419,\"_322\":3424,\"_324\":325},\"e6d6236dd62d\",[3420],{\"_37\":3421,\"_23\":318,\"_319\":3422,\"_174\":3423},\"abe191fa1b690\",[508],\"(It’s also worth noting that with the passage of time, the economics of launching a LLM internally will also likely become more favorable. Specifically, while compute resources are currently scarce and relatively costly, increased commoditization will likely continue to put downward pressure on pricing.)\",[],{\"_37\":3426,\"_23\":313,\"_314\":3427,\"_322\":3432,\"_324\":3178},\"1d054ff63e37\",[3428],{\"_37\":3429,\"_23\":318,\"_319\":3430,\"_174\":3431},\"8b95ba2c10680\",[508],\"Phase 4: Specializing Mature Enterprises Seek Appropriate ML Infrastructure for Use Case Fit\",[],{\"_37\":3434,\"_23\":313,\"_314\":3435,\"_322\":3440,\"_324\":325},\"ff37c886ce95\",[3436],{\"_37\":3437,\"_23\":318,\"_319\":3438,\"_174\":3439},\"8c0686bd91ba0\",[],\"Will all enterprises eventually “graduate” to a state of hosting their own internal, closed-circuit LLM, through which they run every single data job? Not necessarily.\",[],{\"_37\":3442,\"_23\":313,\"_314\":3443,\"_322\":3448,\"_324\":325},\"ef6ab42bda15\",[3444],{\"_37\":3445,\"_23\":318,\"_319\":3446,\"_174\":3447},\"10126fec5aae0\",[],\"As enterprises learn how to run their own ML programs, they will find themselves less bound by hardline goals to spend a certain amount of dollars on AI initiatives or to acquire specific infrastructure configurations, and more focused on finding infrastructure that fits their specific needs and use cases. In some cases, they may find themselves standing up their own internal LLM, complete with their own internal infrastructure hosting configuration and costs, to handle high-priority use cases that require high responsiveness with the utmost privacy. In other cases, as they increase their comfort levels with hosted solutions and find “good enough” data privacy tooling, they may find better value in letting hosted solutions run massive data jobs that might’ve been too costly or time consuming to run internally.\",[],{\"_37\":3450,\"_23\":313,\"_314\":3451,\"_322\":3464,\"_324\":325},\"c5af32f48d1d\",[3452,3456,3460],{\"_37\":3453,\"_23\":318,\"_319\":3454,\"_174\":3455},\"4c9ed7e3f6590\",[],\"Savvy enterprises will understand that, given the incredible speed of change in the AI space, it will be a best practice to seek AI solutions that offer them \",{\"_37\":3457,\"_23\":318,\"_319\":3458,\"_174\":3459},\"4c9ed7e3f6591\",[508],\"optionality \",{\"_37\":3461,\"_23\":318,\"_319\":3462,\"_174\":3463},\"4c9ed7e3f6592\",[],\"in favor of restrictive, long-term commitments with vendor lock-in. Committing to launching an internal LLM may be exactly what an enterprise needs to compete, but it’s a significant investment that will require continuous operational support afterwards. Conversely, signing a contract with a hosted third-party provider eager to get their business may be a much more lightweight alternative.\",[],{\"_37\":3466,\"_23\":313,\"_314\":3467,\"_322\":3496,\"_324\":325},\"c525ae097c65\",[3468,3472,3476,3480,3485,3488,3493],{\"_37\":3469,\"_23\":318,\"_319\":3470,\"_174\":3471},\"c43817777d390\",[],\"As enterprises mature, they’ll begin thinking more deliberately about their AI inference spend, particularly given the high cost of compute. Suffice it to say, there are many aspects of AI infrastructure that \",{\"_37\":3473,\"_23\":318,\"_319\":3474,\"_174\":3475},\"c43817777d391\",[508],\"let\",{\"_37\":3477,\"_23\":318,\"_319\":3478,\"_174\":3479},\"c43817777d392\",[],\" teams expend huge amounts of money, like attempting to train, and repeatedly re-train, an open-weight model using an entire dataset. In the same way that many businesses repatriated their compute infrastructure from owned data centers to cloud providers–then looked on increasing horror as their cloud computing bills skyrocketed–smart enterprises will look for ways to control their spend, such as seeking model training alternatives like \",{\"_37\":3481,\"_23\":318,\"_319\":3482,\"_174\":3484},\"c43817777d393\",[3483,508],\"568c955f1717\",\"model merging\",{\"_37\":3486,\"_23\":318,\"_319\":3487,\"_174\":846},\"c43817777d394\",[],{\"_37\":3489,\"_23\":318,\"_319\":3490,\"_174\":3492},\"c43817777d395\",[3491,508],\"e11882f5736d\",\"mixture-of-experts\",{\"_37\":3494,\"_23\":318,\"_319\":3495,\"_174\":2627},\"c43817777d396\",[],[3497,3499],{\"_37\":3483,\"_23\":49,\"_439\":3498},\"https://www.marktechpost.com/2023/09/27/what-is-model-merging/\",{\"_37\":3491,\"_23\":49,\"_439\":3500},\"https://www.techtarget.com/searchenterpriseai/feature/Mixture-of-experts-models-explained-What-you-need-to-know\",{\"_37\":3502,\"_23\":313,\"_314\":3503,\"_322\":3508,\"_324\":408},\"dafbc222b633\",[3504],{\"_37\":3505,\"_23\":318,\"_319\":3506,\"_174\":3507},\"3d25edef98620\",[],\"How We’ll Address AI’s Infrastructure Challenges Together\",[],{\"_37\":3510,\"_23\":313,\"_314\":3511,\"_322\":3516,\"_324\":325},\"1fdfe783f225\",[3512],{\"_37\":3513,\"_23\":318,\"_319\":3514,\"_174\":3515},\"0331a53110d10\",[],\"In the next 12 months I expect an explosion in new ideas and innovation from practitioners who are actually doing the work to figure this out at scale. At Heavybit, we’re working to identify and elevate the people working in enterprises and startups that are solving these new challenges, creating best practices, and rallying around common and critical challenges. We plan to share our learnings to help accomplish two goals: First, to provide a clearer path for successful AI programs at enterprise scale, and second: To clarify the requirements for enterprise-scale programs to help ambitious AI tooling startups get enterprise-ready.\",[],{\"_37\":3518,\"_23\":313,\"_314\":3519,\"_322\":3532,\"_324\":325},\"7543a6c1cc24\",[3520,3524,3529],{\"_37\":3521,\"_23\":318,\"_319\":3522,\"_174\":3523},\"df298c2e39550\",[508],\"Stay tuned for more updates on how enterprises are tackling new AI and ML challenges like inference hosting. If you’re an AI startup founder looking to get your product ready for enterprise, feel free to \",{\"_37\":3525,\"_23\":318,\"_319\":3526,\"_174\":3528},\"df298c2e39551\",[3527,508],\"5aee85abc7c7\",\"reach out to us\",{\"_37\":3530,\"_23\":318,\"_319\":3531,\"_174\":2627},\"df298c2e39552\",[508],[3533],{\"_37\":3527,\"_23\":49,\"_439\":3534},\"https://heavybit.typeform.com/to/tP7Lh7?typeform-source=EnterpriseReadyAIBlog1\",{\"_37\":3536,\"_23\":313,\"_314\":3537,\"_322\":3542,\"_324\":325},\"3945049d38f9\",[3538],{\"_37\":3539,\"_23\":318,\"_319\":3540,\"_174\":3541},\"a4b1e5b7a84d0\",[],\"\",[],{\"_37\":3544,\"_23\":313,\"_314\":3545,\"_322\":3550,\"_324\":408},\"fe9ab228423e\",[3546],{\"_37\":3547,\"_23\":318,\"_319\":3548,\"_174\":3549},\"351bde9d46af0\",[],\"More Resources\",[],{\"_37\":3552,\"_23\":313,\"_314\":3553,\"_2171\":1957,\"_2172\":2173,\"_322\":3559,\"_324\":325},\"5ec151d357eb\",[3554],{\"_37\":3555,\"_23\":318,\"_319\":3556,\"_174\":3558},\"271ed3657cc60\",[3557],\"8b73d7a32d0f\",\"Article - Enterprise AI Infrastructure: Privacy, Maturity, Resources\",[3560],{\"_37\":3557,\"_23\":49,\"_439\":3561},\"https://www.heavybit.com/library/article/enterprise-ai-infrastructure-privacy-maturity-resources\",{\"_37\":3563,\"_23\":313,\"_314\":3564,\"_2171\":1957,\"_2172\":2173,\"_322\":3570,\"_324\":325},\"a1cbb0d29017\",[3565],{\"_37\":3566,\"_23\":318,\"_319\":3567,\"_174\":3569},\"17832a451f46\",[3568],\"909f3b6d2e33\",\"Article - AI's Hidden Opportunities\",[3571],{\"_37\":3568,\"_23\":49,\"_439\":3572},\"https://www.heavybit.com/library/article/ai-hidden-opportunities-for-software-developers-swyx\",{\"_37\":3574,\"_23\":313,\"_314\":3575,\"_2171\":1957,\"_2172\":2173,\"_322\":3581,\"_324\":325},\"1274395d131a\",[3576],{\"_37\":3577,\"_23\":318,\"_319\":3578,\"_174\":3580},\"e85dd7bb30cd\",[3579],\"52ef233d6032\",\"Article - How to Launch an AI Startup: Lessons from GitHub Copilot\",[3582],{\"_37\":3579,\"_23\":49,\"_439\":3583},\"https://www.heavybit.com/library/article/how-to-launch-an-ai-startup-github-copilot-lessons\",{\"_37\":3585,\"_23\":313,\"_314\":3586,\"_2171\":1957,\"_2172\":2173,\"_322\":3592,\"_324\":325},\"92cff2edb73e\",[3587],{\"_37\":3588,\"_23\":318,\"_319\":3589,\"_174\":3591},\"e14fc7d995a80\",[3590],\"536cbe87e3b9\",\"Article - Digging Deeper into Building on LLMs, AI Coding Assistants, and Observability\",[3593],{\"_37\":3590,\"_23\":49,\"_439\":2889},{\"_37\":3595,\"_23\":313,\"_314\":3596,\"_2171\":1957,\"_2172\":2173,\"_322\":3602,\"_324\":325},\"6b616a744c4d\",[3597],{\"_37\":3598,\"_23\":318,\"_319\":3599,\"_174\":3601},\"92fe01c6d0770\",[3600],\"7801f3c4dd99\",\"Article - MLOps vs. Eng: Misaligned Incentives and Failure to Launch?\",[3603],{\"_37\":3600,\"_23\":49,\"_439\":631},{\"_37\":3605,\"_23\":313,\"_314\":3606,\"_2171\":1957,\"_2172\":2173,\"_322\":3612,\"_324\":325},\"b2d2ae9b38b7\",[3607],{\"_37\":3608,\"_23\":318,\"_319\":3609,\"_174\":3611},\"2d328a22542e\",[3610],\"d1ec9d95b973\",\"Article - Incident Response and DevOps in the Age of Generative AI\",[3613],{\"_37\":3610,\"_23\":49,\"_439\":3614},\"https://www.heavybit.com/library/article/generative-ai-incident-response-devops\",17,\"https://cdn.sanity.io/images/50q6fr1p/production/364de80e162115d62cf78c2c8581e36ac77e363f-1200x630.png\",\"objectID\",\"3dbfba3b-523f-4ec6-b0fc-17a26f0f9ec3\",\"2024-09-16T16:13:00.000Z\",\"ai-infrastructure-top-challenges-data-inference\",\"typeName\",{\"_398\":3623,\"_2820\":3675,\"_149\":3676,\"_3617\":3677,\"_3678\":3679,\"_912\":3680,\"_366\":3681,\"_32\":3682,\"_3621\":3683},[3624,3632,3646,3655,3664],{\"_37\":3625,\"_23\":313,\"_314\":3626,\"_322\":3631,\"_324\":325},\"c6526ecf9611\",[3627],{\"_37\":3628,\"_23\":318,\"_319\":3629,\"_174\":3630},\"c899e83356e00\",[],\"In this panel we'll dig into the current state of open source licensing and what is and is not open source, what the current trends around BSL mean for the future of open source communities and companies alike, and finally how generative AI tools like ChatGPT, Copilot, Bard, and others are disrupting the OSS landscape, for better or worse.\",[],{\"_37\":3633,\"_23\":313,\"_314\":3634,\"_322\":3639,\"_324\":325},\"8f0c73212134\",[3635],{\"_37\":3636,\"_23\":318,\"_319\":3637,\"_174\":3638},\"8e1b2c6392bf\",[508],\"Related articles: \",[3640,3643],{\"_37\":3641,\"_23\":49,\"_439\":3642},\"5269c0e18210\",\"https://www.heavybit.com/library/article/open-source-software-benefits-advantages\",{\"_37\":3644,\"_23\":49,\"_439\":3645},\"a139d1b98986\",\"https://www.heavybit.com/library/article/open-source-vs-proprietary\",{\"_37\":3647,\"_23\":313,\"_314\":3648,\"_2171\":1957,\"_2172\":2173,\"_322\":3653,\"_324\":325},\"b36fd0fcdaa8\",[3649],{\"_37\":3650,\"_23\":318,\"_319\":3651,\"_174\":3652},\"6902b5f0ca32\",[508,3641],\"Advantages of Open-Source Software\",[3654],{\"_37\":3641,\"_23\":49,\"_439\":3642},{\"_37\":3656,\"_23\":313,\"_314\":3657,\"_2171\":1957,\"_2172\":2173,\"_322\":3662,\"_324\":325},\"115d834638fb\",[3658],{\"_37\":3659,\"_23\":318,\"_319\":3660,\"_174\":3661},\"d8d442f1e656\",[508,3644],\"Differences Between Proprietary and Open-Source Software\",[3663],{\"_37\":3644,\"_23\":49,\"_439\":3645},{\"_37\":3665,\"_23\":313,\"_314\":3666,\"_2171\":1957,\"_2172\":2173,\"_322\":3672,\"_324\":325},\"0c77af81b214\",[3667],{\"_37\":3668,\"_23\":318,\"_319\":3669,\"_174\":3671},\"6dc48c609b14\",[3670,508],\"4ae63d3d6ccd\",\"How to Start an Open-Source Project\",[3673],{\"_37\":3670,\"_23\":49,\"_439\":3674},\"https://www.heavybit.com/library/article/how-to-start-an-open-source-project\",1916.375089,\"https://image.mux.com/sUGdbEvftGxgJ2a8x01EecSZIQwMjvshO/thumbnail.png\",\"35ac10d9-593e-4a7e-8fb6-abf47ccc3d8b\",\"poster\",\"mux\",\"2023-06-29T18:32:31.818Z\",\"open-source-licensing-and-the-future-of-open-source-businesses\",\"Open-Source Licensing and The Future of Open Source Businesses\",\"video\",{\"_398\":3685,\"_2820\":4110,\"_149\":4111,\"_3617\":4112,\"_912\":4113,\"_366\":4114,\"_32\":4115,\"_3621\":278},[3686,3694,3702,3749,3757,3761,3769,3777,3806,3814,3822,3830,3838,3846,3854,3888,3900,3912,3924,3936,3944,3952,3960,3968,3976,3984,3992,4022,4030,4038,4046,4054,4062,4070,4078,4086,4094,4102],{\"_37\":3687,\"_23\":313,\"_314\":3688,\"_322\":3693,\"_324\":408},\"9196fa5c1337\",[3689],{\"_37\":3690,\"_23\":318,\"_319\":3691,\"_174\":3692},\"daa6967db340\",[],\"The Unique Challenges of Mobile Compute\",[],{\"_37\":3695,\"_23\":313,\"_314\":3696,\"_322\":3701,\"_324\":325},\"76f45bfec570\",[3697],{\"_37\":3698,\"_23\":318,\"_319\":3699,\"_174\":3700},\"6c276d156142\",[],\"A significant focus in modern AI has been on large language models with billions of parameters. Billions more parameters mean bigger training investment, and presumably, more-impressive performance. But there are emerging use cases for mobile AI inference, like offline mobile text generation and the rise of general-purpose generative models for edge that hint at a growing need for reliable mobile AI performance.\",[],{\"_37\":3703,\"_23\":313,\"_314\":3704,\"_322\":3740,\"_324\":325},\"af0f0de9eaae\",[3705,3710,3714,3719,3722,3727,3731,3736],{\"_37\":3706,\"_23\":318,\"_319\":3707,\"_174\":3709},\"bce9edf18932\",[3708],\"01f586bb19c8\",\"Chris Karani\",{\"_37\":3711,\"_23\":318,\"_319\":3712,\"_174\":3713},\"669b28ad0b26\",[],\", a longtime software engineer with specific expertise working in the mobile space, has built a variety of open-source, mobile-specific AI projects like \",{\"_37\":3715,\"_23\":318,\"_319\":3716,\"_174\":3718},\"8b6f89ec9c39\",[3717],\"dc8c3118f86c\",\"Wax\",{\"_37\":3720,\"_23\":318,\"_319\":3721,\"_174\":846},\"43437631fed2\",[],{\"_37\":3723,\"_23\":318,\"_319\":3724,\"_174\":3726},\"4fbb11a36374\",[3725],\"4edd5928711b\",\"Swarm\",{\"_37\":3728,\"_23\":318,\"_319\":3729,\"_174\":3730},\"e59bd93b5314\",[],\", and is now focusing on building a mobile-specific monitoring and release-gating platform \",{\"_37\":3732,\"_23\":318,\"_319\":3733,\"_174\":3735},\"aa0cafd1ebca\",[3734],\"a7fbf0b65b8d\",\"RYNO\",{\"_37\":3737,\"_23\":318,\"_319\":3738,\"_174\":3739},\"4ab51806f777\",[],\" after observing the unique challenges that on-device inference poses to software performance on mobile.\",[3741,3743,3745,3747],{\"_37\":3708,\"_23\":49,\"_439\":3742},\"https://www.linkedin.com/in/chriskarani/\",{\"_37\":3717,\"_23\":49,\"_439\":3744},\"https://github.com/christopherkarani/Wax\",{\"_37\":3725,\"_23\":49,\"_439\":3746},\"https://github.com/christopherkarani/Swarm\",{\"_37\":3734,\"_23\":49,\"_439\":3748},\"https://www.youtube.com/watch?v=yL9lcmvSs7o\",{\"_37\":3750,\"_23\":313,\"_314\":3751,\"_322\":3756,\"_324\":325},\"6a7d245a9398\",[3752],{\"_37\":3753,\"_23\":318,\"_319\":3754,\"_174\":3755},\"d215b9275145\",[],\"Below, he explains why mobile inference needs a new breed of observability, and why on-device inference may necessitate a different way to think about app development: At the edge, moving inference on-device effectively introduces a variety of new hardware-specific variables, such as device generations, OS versions, and thermal and memory constraints.\",[],{\"_37\":3758,\"_23\":149,\"_114\":3759},\"68822d00ffc2\",{\"_116\":3760,\"_23\":118},\"image-1dd5e966858d3cb95946cf1b3d940b1c92f2c46a-2030x554-png\",{\"_37\":3762,\"_23\":313,\"_314\":3763,\"_322\":3768,\"_324\":325},\"23fa169cb01d\",[3764],{\"_37\":3765,\"_23\":318,\"_319\":3766,\"_174\":3767},\"57f6112d5429\",[508],\"The RYNO dashboard specifically monitors on-device inference.\",[],{\"_37\":3770,\"_23\":313,\"_314\":3771,\"_322\":3776,\"_324\":408},\"5d05e23e9ffc\",[3772],{\"_37\":3773,\"_23\":318,\"_319\":3774,\"_174\":3775},\"d169649fc0a1\",[],\"How Traditional Observability Can’t Cover AI at the Edge\",[],{\"_37\":3778,\"_23\":313,\"_314\":3779,\"_322\":3801,\"_324\":325},\"1e42fff1e07f\",[3780,3784,3789,3792,3797],{\"_37\":3781,\"_23\":318,\"_319\":3782,\"_174\":3783},\"c4f86ecd4a57\",[],\"Karani explains that as a builder in the mobile space, he initially relied on popular tools like \",{\"_37\":3785,\"_23\":318,\"_319\":3786,\"_174\":3788},\"471e6fe7b6f2\",[3787],\"f7f739690c4b\",\"LangFuse\",{\"_37\":3790,\"_23\":318,\"_319\":3791,\"_174\":846},\"4d633666c652\",[],{\"_37\":3793,\"_23\":318,\"_319\":3794,\"_174\":3796},\"a54412f6e2d3\",[3795],\"f42d14678a11\",\"LangSmith\",{\"_37\":3798,\"_23\":318,\"_319\":3799,\"_174\":3800},\"ad6e4e5ad7b2\",[],\" for observability, but found that neither provided the monitoring coverage his projects needed. He found himself building RYNO, a flight recorder and release gate for AI in production on mobile, which focuses on privacy, safe execution traces from real devices, and turning production failures into regression tests to improve future releases.\",[3802,3804],{\"_37\":3787,\"_23\":49,\"_439\":3803},\"https://langfuse.com/\",{\"_37\":3795,\"_23\":49,\"_439\":3805},\"https://www.langchain.com/langsmith-platform\",{\"_37\":3807,\"_23\":313,\"_314\":3808,\"_322\":3813,\"_324\":325},\"4a79b7587c7b\",[3809],{\"_37\":3810,\"_23\":318,\"_319\":3811,\"_174\":3812},\"f0af90a0db90\",[],\"“The project combines the hardware state with the end AI behavior to let teams debug and figure out why model failures happen when the apps are actually in production,” the founder explains. As an example, Karani cites a team he spoke with whose mobile app that identifies plants out in the wild and diagnoses them for issues.\",[],{\"_37\":3815,\"_23\":313,\"_314\":3816,\"_322\":3821,\"_324\":325},\"f8c80b58c7b7\",[3817],{\"_37\":3818,\"_23\":318,\"_319\":3819,\"_174\":3820},\"756f14956665\",[],\"Over time, the app team noticed negative reviews in the App Store from frustrated users who found the app misidentifying plants, but when the app’s developers conducted the same tests internally, they noticed no errors. “What they didn't realize was that a lot of their users were out using their apps in high ambient temperatures. This causes a variety of issues that the existing tools just couldn't catch.”\",[],{\"_37\":3823,\"_23\":313,\"_314\":3824,\"_322\":3829,\"_324\":325},\"73f032e4b7e8\",[3825],{\"_37\":3826,\"_23\":318,\"_319\":3827,\"_174\":3828},\"4a033185569e\",[],\"“For instance, once inference is running on a phone in production, changes in thermal state, memory pressure, and available compute can change how that workload performs in production. You can see slower execution, memory-related failures, or different behavior from what you saw during testing. RYNO comes in and is able to combine the hardware state with the model output and help teams debug why their models are failing when they're out into production.”\",[],{\"_37\":3831,\"_23\":313,\"_314\":3832,\"_322\":3837,\"_324\":408},\"cc837e4097a6\",[3833],{\"_37\":3834,\"_23\":318,\"_319\":3835,\"_174\":3836},\"4be7b5ebf70f\",[],\"Closing the Loop for Mobile: Ragequitting, UX, and Regressions\",[],{\"_37\":3839,\"_23\":313,\"_314\":3840,\"_322\":3845,\"_324\":325},\"ac1a98cd7c62\",[3841],{\"_37\":3842,\"_23\":318,\"_319\":3843,\"_174\":3844},\"cb6169cdaa7e\",[],\"For consumer-level mobile applications, this type of failure can lead to the ultimate cost: Users ragequitting the app and immediately deleting it. While the founder’s aim is eventually to target optimization opportunities for any edge computing setup, he admits that mobile is his current focus. “In mobile, most apps have a high uninstall rate.”\",[],{\"_37\":3847,\"_23\":313,\"_314\":3848,\"_322\":3853,\"_324\":325},\"09e3da705404\",[3849],{\"_37\":3850,\"_23\":318,\"_319\":3851,\"_174\":3852},\"bb84bfa9efca\",[],\"To avoid grossly inaccurate outputs due to hallucinations, Karani explains that his platform runs AI jobs across five stages: Capture, explain, replay, promote, and release.\",[],{\"_37\":3855,\"_23\":313,\"_314\":3856,\"_2171\":1957,\"_2172\":2173,\"_322\":3883,\"_324\":325},\"a3c417f4e89a\",[3857,3861,3865,3870,3874,3879],{\"_37\":3858,\"_23\":318,\"_319\":3859,\"_174\":3860},\"14fdc7cba99a\",[2165],\"Capture:\",{\"_37\":3862,\"_23\":318,\"_319\":3863,\"_174\":3864},\"aa00c58d638c\",[],\" Utilizing the open-source, lightweight, privacy-centric \",{\"_37\":3866,\"_23\":318,\"_319\":3867,\"_174\":3869},\"62c5efc259aa\",[3868],\"39f4350f1285\",\"Terra SDK\",{\"_37\":3871,\"_23\":318,\"_319\":3872,\"_174\":3873},\"faaca851a04b\",[],\", which can be embedded into mobile apps, and is built on \",{\"_37\":3875,\"_23\":318,\"_319\":3876,\"_174\":3878},\"2b5210b4a320\",[3877],\"12ed0e79639f\",\"open telemetry\",{\"_37\":3880,\"_23\":318,\"_319\":3881,\"_174\":3882},\"1cf657c9feb9\",[],\" to prevent vendor lock-in, and records OS version, memory pressure, thermal state, and records OS version, memory pressure, thermal state, device class, and the compute resources involved in inference, such as CPU, GPU, or Neural Engine where available\",[3884,3886],{\"_37\":3868,\"_23\":49,\"_439\":3885},\"https://github.com/tryterra/terra-react\",{\"_37\":3877,\"_23\":49,\"_439\":3887},\"https://opentelemetry.io\",{\"_37\":3889,\"_23\":313,\"_314\":3890,\"_2171\":1957,\"_2172\":2173,\"_322\":3899,\"_324\":325},\"6bfe8b8aa55b\",[3891,3895],{\"_37\":3892,\"_23\":318,\"_319\":3893,\"_174\":3894},\"531d8bfb8a6c\",[2165],\"Explain:\",{\"_37\":3896,\"_23\":318,\"_319\":3897,\"_174\":3898},\"102c1b804637\",[],\" Utilizing the local macOS app’s ability to plot a visual timeline of device physics and model execution, identifying factors like correlating changes in device state with changes in model execution\",[],{\"_37\":3901,\"_23\":313,\"_314\":3902,\"_2171\":1957,\"_2172\":2173,\"_322\":3911,\"_324\":325},\"16a216eace7b\",[3903,3907],{\"_37\":3904,\"_23\":318,\"_319\":3905,\"_174\":3906},\"005091edab91\",[2165],\"Replay:\",{\"_37\":3908,\"_23\":318,\"_319\":3909,\"_174\":3910},\"6e1f8d5da353\",[],\" Having captured structured metadata, template IDs, token counts, and the precise physical hardware context at the time of errors, developers can use a replay button to recreate the conditions surrounding the failure to artificially reproduce the same environmental stress for testing purposes\",[],{\"_37\":3913,\"_23\":313,\"_314\":3914,\"_2171\":1957,\"_2172\":2173,\"_322\":3923,\"_324\":325},\"64a94ff32346\",[3915,3919],{\"_37\":3916,\"_23\":318,\"_319\":3917,\"_174\":3918},\"c2cb193596ec\",[2165],\"Promote:\",{\"_37\":3920,\"_23\":318,\"_319\":3921,\"_174\":3922},\"057867fa95b8\",[],\" Teams can turn a one-off production failure into a structured, repeatable evaluation\",[],{\"_37\":3925,\"_23\":313,\"_314\":3926,\"_2171\":1957,\"_2172\":2173,\"_322\":3935,\"_324\":325},\"523186127ddd\",[3927,3931],{\"_37\":3928,\"_23\":318,\"_319\":3929,\"_174\":3930},\"3f0388696f60\",[2165],\"Release:\",{\"_37\":3932,\"_23\":318,\"_319\":3933,\"_174\":3934},\"3802f3236628\",[],\" From here, the platform connects to a team CI/CD pipeline to run various evaluations on different devices with that specific eval case prior to the next app release to prevent further regressions before they make it to production\",[],{\"_37\":3937,\"_23\":313,\"_314\":3938,\"_322\":3943,\"_324\":325},\"88dd59afe227\",[3939],{\"_37\":3940,\"_23\":318,\"_319\":3941,\"_174\":3942},\"6cf15725e554\",[],\"The founder adds: “A lot of these issues also come in simple OS updates. Everything was working fine, then Apple launches a new update that introduces unexpected changes. And sometimes getting that stress test when you're testing within a controlled environment is difficult. So a lot of these issues can only be caught when the AI is actually ‘out in the wild.’”\",[],{\"_37\":3945,\"_23\":313,\"_314\":3946,\"_322\":3951,\"_324\":325},\"a4d61e87b83e\",[3947],{\"_37\":3948,\"_23\":318,\"_319\":3949,\"_174\":3950},\"c63132c0118c\",[],\"“Sometimes you have to actually catch these failures once they're actually already out there because these systems are so non deterministic. They're not like traditional, deterministic software where you might have a really tight unit test case, so certain errors don’t reach production.”\",[],{\"_37\":3953,\"_23\":313,\"_314\":3954,\"_322\":3959,\"_324\":325},\"51f37cb77d65\",[3955],{\"_37\":3956,\"_23\":318,\"_319\":3957,\"_174\":3958},\"9098f81dfbb7\",[],\"“Because there's 100 million ways that the AI can actually fail. So, one of the best approaches is having this sort of loop that we've created to catch them quickly in production and ensure that they don't persist in the future.”\",[],{\"_37\":3961,\"_23\":313,\"_314\":3962,\"_322\":3967,\"_324\":408},\"6e814b0ff721\",[3963],{\"_37\":3964,\"_23\":318,\"_319\":3965,\"_174\":3966},\"07bb1c6dbf0f\",[],\"Building for Privacy and Memory Optimization\",[],{\"_37\":3969,\"_23\":313,\"_314\":3970,\"_322\":3975,\"_324\":325},\"564424a3cff6\",[3971],{\"_37\":3972,\"_23\":318,\"_319\":3973,\"_174\":3974},\"4c7d752edc34\",[],\"Karani explains that privacy was a key part of the project due to its initial focus on iOS. “We believe that the on-device AI on iOS is some of the best in the world right now. The Apple Neural Engine lets users run these really awesome models. But for iOS and mobile users, privacy is very important. It’s common even for experienced teams to accidentally violate user privacy while trying to collect the debugging context they desperately need.”\",[],{\"_37\":3977,\"_23\":313,\"_314\":3978,\"_322\":3983,\"_324\":325},\"1d35f785606e\",[3979],{\"_37\":3980,\"_23\":318,\"_319\":3981,\"_174\":3982},\"8a890a6f2be8\",[],\"“This is how teams are silently trying to capture model regression with internal tooling. They captured a prompt and its output, and those contain extremely sensitive data, so you end up violating user privacy. To fix that for RYNO, we actually have a ‘never log’ list. No screenshots, no user prompts, no contacts lists, no physical addresses, no chat transcripts.”\",[],{\"_37\":3985,\"_23\":313,\"_314\":3986,\"_322\":3991,\"_324\":325},\"846ca4d16b40\",[3987],{\"_37\":3988,\"_23\":318,\"_319\":3989,\"_174\":3990},\"8da126fcdaa0\",[],\"“We use a minimal telemetry schema. Instead of logging what the user has typed, we log a structured shape and the operational health of the mathematical transaction. We've been using a lot of tools to ensure that teams are able to get the metadata that they need to reproduce these failures while still avoiding violations of user privacy in the process. We think this is super-important for on-device AI.”\",[],{\"_37\":3993,\"_23\":313,\"_314\":3994,\"_322\":4017,\"_324\":325},\"ea0d0888f24d\",[3995,3999,4004,4008,4013],{\"_37\":3996,\"_23\":318,\"_319\":3997,\"_174\":3998},\"c484d592a241\",[],\"Aside from privacy, the founder notes that going lightweight was an important architectural consideration to run successfully on edge. “Because of the memory pressure issues with the \",{\"_37\":4000,\"_23\":318,\"_319\":4001,\"_174\":4003},\"e39d37c1ea2c\",[4002],\"a281272b15f2\",\"Jetson Linux\",{\"_37\":4005,\"_23\":318,\"_319\":4006,\"_174\":4007},\"7c27af004f44\",[],\", we made a key decision early on to build the core of the SDK in \",{\"_37\":4009,\"_23\":318,\"_319\":4010,\"_174\":4012},\"a74babccffb6\",[4011],\"a0ea04ccc546\",\"Zig\",{\"_37\":4014,\"_23\":318,\"_319\":4015,\"_174\":4016},\"49524083c6a5\",[],\". This gave us a lot of control over memory pressure and it gave us full control over how we actually want to architect memory use for the SDK, because on-device AI uses a lot of resources on mobile.”\",[4018,4020],{\"_37\":4002,\"_23\":49,\"_439\":4019},\"https://developer.nvidia.com/embedded/jetson-linux-r3643\",{\"_37\":4011,\"_23\":49,\"_439\":4021},\"https://ziglang.org/\",{\"_37\":4023,\"_23\":313,\"_314\":4024,\"_322\":4029,\"_324\":325},\"e379b830e1eb\",[4025],{\"_37\":4026,\"_23\":318,\"_319\":4027,\"_174\":4028},\"090dfbc5e2fc\",[],\"“So we needed to build an SDK that's extremely lightweight. Because we’re running an observability SDK on constrained hardware, it would need to stay lightweight enough to not create another source of memory pressure competing against the model it’s observing. We think that for edge AI, debugging is something that is going to become a big problem. And it's totally different from how current tools actually enable teams to receive telemetry.”\",[],{\"_37\":4031,\"_23\":313,\"_314\":4032,\"_322\":4037,\"_324\":325},\"d120d7d1c55e\",[4033],{\"_37\":4034,\"_23\":318,\"_319\":4035,\"_174\":4036},\"017f00a9c3a6\",[],\"The founder admits that part of his inspiration came from his own analysis of the AI space, which increasingly seems dominated by billion-dollar vendors pushing trillion-parameter models. “We believe wholeheartedly that edge AI is 100% the future. Smaller models are improving at a rapid rate. Vendors like Apple and Google keep pumping a lot of money in the space, and we think that the kind of benefits that edge AI provides are undeniable.”\",[],{\"_37\":4039,\"_23\":313,\"_314\":4040,\"_322\":4045,\"_324\":325},\"88259a07d188\",[4041],{\"_37\":4042,\"_23\":318,\"_319\":4043,\"_174\":4044},\"007257714f01\",[],\"“First, there’s privacy. Second, there’s improvements to latency, and third, is cost reduction. As the models keep improving, people are going to start realizing and taking this extremely seriously as we move forward. The space is still very early. But the world I see in the future is one in which we'll have on-device AI everywhere you go. There’s already on-device AI in cars, and we'll have robots that need on-device AI, for television, mobile, and so on.”\",[],{\"_37\":4047,\"_23\":313,\"_314\":4048,\"_322\":4053,\"_324\":325},\"fa06d141f9b1\",[4049],{\"_37\":4050,\"_23\":318,\"_319\":4051,\"_174\":4052},\"8c15a64289b3\",[],\"“Yes, the Cloud will always be where true intelligence, the larger computational intelligence, lives. But more and more companies are going to keep moving towards leveraging the on-device AI space as we move forward. And I think that's why Apple has also been putting a lot of effort year over year into the Apple Neural Engine and every chip that they release, prioritizing the inference power that they can actually pull out of the on device models.”\",[],{\"_37\":4055,\"_23\":313,\"_314\":4056,\"_322\":4061,\"_324\":408},\"bb08c1e33c67\",[4057],{\"_37\":4058,\"_23\":318,\"_319\":4059,\"_174\":4060},\"8225a1eccfb5\",[],\"Founder Lessons, Enterprise Adoption, and the Future of Compute\",[],{\"_37\":4063,\"_23\":313,\"_314\":4064,\"_322\":4069,\"_324\":325},\"0f1dca34a263\",[4065],{\"_37\":4066,\"_23\":318,\"_319\":4067,\"_174\":4068},\"5446dd4a7a6b\",[],\"For founders looking to build on-device AI products, Karani suggests that his own deep understanding of the space has been fundamental to his progress. “I think having a good understanding of how the hardware state actually works is important. The computation and the OS system attributes during inference time...that’s a different problem space that a lot of engineers need to understand if they're going into the on-device space.”\",[],{\"_37\":4071,\"_23\":313,\"_314\":4072,\"_322\":4077,\"_324\":325},\"2f49784b8a7f\",[4073],{\"_37\":4074,\"_23\":318,\"_319\":4075,\"_174\":4076},\"bd75e9744e7c\",[],\"“Inference is one thing, input and output of prompts is another. But for a lot of these models, you have to understand they're not in a problem space that is similar to running on a large data center in Atlanta with perfect resources all the time. These are models that are running on-device with varying, different pressures in constrained environments. Understanding those environments is key to actually building the right systems for AI solutions.”\",[],{\"_37\":4079,\"_23\":313,\"_314\":4080,\"_322\":4085,\"_324\":325},\"d217ab50b70d\",[4081],{\"_37\":4082,\"_23\":318,\"_319\":4083,\"_174\":4084},\"80ee14ec9ff8\",[],\"On the question of why device-specific telemetry still doesn’t seem to be mainstream at the enterprise level, the founder points back to how the tool ecosystem for edge computing is still in its early days. “There’s a lack of tooling. A lot of the enterprise companies we speak to are concerned over their lack of telemetry and understanding of how models behave when they’re actually out into production.”\",[],{\"_37\":4087,\"_23\":313,\"_314\":4088,\"_322\":4093,\"_324\":325},\"66d1e4b93387\",[4089],{\"_37\":4090,\"_23\":318,\"_319\":4091,\"_174\":4092},\"f3f842438cd8\",[],\"“For instance, we were speaking to a medical company that was trying to launch a very privacy-centric, on-device model to different customers. But what they found is that they just couldn’t risk getting an incorrect diagnosis from the model (and not being able to understand why it happened). Existing tools just didn't cater to them. The space is so early and emerging and a lot of the focus is on cloud models.”\",[],{\"_37\":4095,\"_23\":313,\"_314\":4096,\"_322\":4101,\"_324\":325},\"37063324755f\",[4097],{\"_37\":4098,\"_23\":318,\"_319\":4099,\"_174\":4100},\"345027e2fcbe\",[],\"The founder notes that while high-end cloud models get a lot of attention, they’re also currently leading to a lot of untenable costs. “I think that's what's going to push enterprises more into on-device AI. And as [on-device models] improve, the solutions will become simpler, but we need to provide them with the kind of tooling that enables them to ship on-device AI reliably to their users.”\",[],{\"_37\":4103,\"_23\":313,\"_314\":4104,\"_322\":4109,\"_324\":325},\"8e071c2e06f0\",[4105],{\"_37\":4106,\"_23\":318,\"_319\":4107,\"_174\":4108},\"6731a73112cc\",[],\"Will there be an inflection point at which everything flips from cloud to local? Maybe not. “I think people are going to have a balance of the two. Teams will leverage on-device models for certain tasks and route to cloud models when they need heavy computational work to be completed. I think the future will be a hybrid of the two. I don’t think on-device AI is going to completely negate the need for Cloud, but I think cost and performance are going to push people to try to understand how to leverage a hybrid of the two. And the majority of the work we do today will be pushed to the on-device models for inference.”\",[],13,\"https://cdn.sanity.io/images/50q6fr1p/production/b4042cb54ee8f7b7f4ab04864bd8e1cf7a157ff6-1200x630.png\",\"f7ca2bf6-bf0c-4904-aa01-faede655849e\",\"2026-08-25T16:22:00.000Z\",\"why-on-device-inference-needs-custom-observability\",\"Why On-Device Inference Needs Custom Observability\",\"recent\",[4118,4726,4973],{\"_398\":4119,\"_2820\":3615,\"_149\":4721,\"_3617\":4722,\"_912\":4723,\"_366\":4724,\"_32\":4725,\"_3621\":278},[4120,4128,4158,4187,4191,4206,4214,4233,4252,4260,4268,4276,4306,4314,4322,4330,4338,4357,4376,4384,4392,4419,4437,4453,4461,4469,4477,4485,4493,4501,4531,4539,4547,4555,4563,4571,4579,4587,4627,4657,4665,4673,4689,4697,4713],{\"_37\":4121,\"_23\":313,\"_314\":4122,\"_322\":4127,\"_324\":408},\"f09e9612ef15\",[4123],{\"_37\":4124,\"_23\":318,\"_319\":4125,\"_174\":4126},\"85de00f41251\",[],\"Why Code-as-Text Doesn’t Work for Genuine Understanding\",[],{\"_37\":4129,\"_23\":313,\"_314\":4130,\"_322\":4153,\"_324\":325},\"4b5755f9b8bc\",[4131,4135,4140,4144,4149],{\"_37\":4132,\"_23\":318,\"_319\":4133,\"_174\":4134},\"adc4d52ad4cd\",[],\"Major engineering orgs claim that as much as \",{\"_37\":4136,\"_23\":318,\"_319\":4137,\"_174\":4139},\"d1aebf2378c0\",[4138],\"afe823ab102f\",\"30%\",{\"_37\":4141,\"_23\":318,\"_319\":4142,\"_174\":4143},\"47cf0e8a331c\",[],\" to \",{\"_37\":4145,\"_23\":318,\"_319\":4146,\"_174\":4148},\"a7022507758b\",[4147],\"8042bdcd04fd\",\"75%\",{\"_37\":4150,\"_23\":318,\"_319\":4151,\"_174\":4152},\"70dd8d1c5fc1\",[],\" of their new code is AI-generated, which does nothing to help the ongoing challenges of first-time onboarding for codebases and the updating of existing codebases for newly-hired engineers and growing teams (as well as for agentic coding assistants).\",[4154,4156],{\"_37\":4138,\"_23\":49,\"_439\":4155},\"https://techcrunch.com/2025/04/29/microsoft-ceo-says-up-to-30-of-the-companys-code-was-written-by-ai/\",{\"_37\":4147,\"_23\":49,\"_439\":4157},\"https://blog.google/innovation-and-ai/infrastructure-and-cloud/google-cloud/cloud-next-2026-sundar-pichai/\",{\"_37\":4159,\"_23\":313,\"_314\":4160,\"_322\":4182,\"_324\":325},\"9b432548c002\",[4161,4165,4170,4174,4179],{\"_37\":4162,\"_23\":318,\"_319\":4163,\"_174\":4164},\"9c50f1246f14\",[],\"As generating code becomes free, understanding what any of it does becomes much more costly, and updating it becomes significantly more complex. Developer \",{\"_37\":4166,\"_23\":318,\"_319\":4167,\"_174\":4169},\"7aa3b54f64f6\",[4168],\"a50030cc5b04\",\"Vitali Avagyan\",{\"_37\":4171,\"_23\":318,\"_319\":4172,\"_174\":4173},\"a0f94686e2c4\",[],\" asserts that the key to understanding code, particularly code across different languages, is to stop thinking of it as lines of plain text, and to treat it as a relational graph, which is the basis of his open-source project \",{\"_37\":4175,\"_23\":318,\"_319\":4176,\"_174\":4178},\"c66f9ef7a776\",[4177],\"b6ce27c00ce8\",\"code-graph-rag\",{\"_37\":4180,\"_23\":318,\"_319\":4181,\"_174\":2627},\"c24ce7265e1d\",[],[4183,4185],{\"_37\":4168,\"_23\":49,\"_439\":4184},\"https://www.linkedin.com/in/vitali-avagyan-a1566234/\",{\"_37\":4177,\"_23\":49,\"_439\":4186},\"https://github.com/vitali87/code-graph-rag\",{\"_37\":4188,\"_23\":149,\"_114\":4189},\"c9d37b6922a5\",{\"_116\":4190,\"_23\":118},\"image-bc24d7b1e62740bb7448a1a80a9dfd59940c68d6-3840x1100-jpg\",{\"_37\":4192,\"_23\":313,\"_314\":4193,\"_322\":4203,\"_324\":325},\"b80310699127\",[4194,4198],{\"_37\":4195,\"_23\":318,\"_319\":4196,\"_174\":4197},\"a0931bc06a15\",[508],\"Vitali Avagyan discusses using GraphRAG to parse repos with Memgraph. Image courtesy \",{\"_37\":4199,\"_23\":318,\"_319\":4200,\"_174\":4202},\"5ef047d4ccf9\",[4201,508],\"5b72c578e5d1\",\"Memgraph\",[4204],{\"_37\":4201,\"_23\":49,\"_439\":4205},\"https://www.youtube.com/watch?v=O2jUTq6nCEY\",{\"_37\":4207,\"_23\":313,\"_314\":4208,\"_322\":4213,\"_324\":408},\"2fc0e81b9ef7\",[4209],{\"_37\":4210,\"_23\":318,\"_319\":4211,\"_174\":4212},\"4cb4e6406d24\",[],\"Viewing Code as a Retrievable Graph\",[],{\"_37\":4215,\"_23\":313,\"_314\":4216,\"_322\":4230,\"_324\":325},\"c357da4c0347\",[4217,4221,4226],{\"_37\":4218,\"_23\":318,\"_319\":4219,\"_174\":4220},\"b42b4a2c8abb\",[],\"Avagyan notes that when you hand a codebase to one of today’s coding assistants, it’ll often treat it as a text artifact. “Then, it traverses the code base with a \",{\"_37\":4222,\"_23\":318,\"_319\":4223,\"_174\":4225},\"683e39e682aa\",[4224],\"67f4fbd2fe9a\",\"grep\",{\"_37\":4227,\"_23\":318,\"_319\":4228,\"_174\":4229},\"7de49dd818b7\",[],\" or possibly some kind of agentic search. As you might imagine, that can become inefficient, especially considering some of the questions you might typically have about a codebase.”\",[4231],{\"_37\":4224,\"_23\":49,\"_439\":4232},\"https://en.wikipedia.org/wiki/Grep\",{\"_37\":4234,\"_23\":313,\"_314\":4235,\"_322\":4249,\"_324\":325},\"6bfeff21d7b3\",[4236,4240,4245],{\"_37\":4237,\"_23\":318,\"_319\":4238,\"_174\":4239},\"2814a2a422f6\",[],\"The creator gives the example of asking an agent to find the longest nested function call in your codebase. Today’s agents will search through the imports and invocations and do some counting, a task that LLMs aren’t great at. “But if you already have a graph that represents all core relationships between methods and functions, you or an AI agent can just write a \",{\"_37\":4241,\"_23\":318,\"_319\":4242,\"_174\":4244},\"fd7434afa3fb\",[4243],\"40d5a9d1fab0\",\"Cypher\",{\"_37\":4246,\"_23\":318,\"_319\":4247,\"_174\":4248},\"850209620fae\",[],\" query that will calculate it for you in milliseconds.” The creator suggests that treating codebases as graphs in this way can conserve both tokens and time.\",[4250],{\"_37\":4243,\"_23\":49,\"_439\":4251},\"https://en.wikipedia.org/wiki/Cypher_(query_language)\",{\"_37\":4253,\"_23\":313,\"_314\":4254,\"_322\":4259,\"_324\":325},\"8edb1e48b5a2\",[4255],{\"_37\":4256,\"_23\":318,\"_319\":4257,\"_174\":4258},\"b5cb7e6813d4\",[],\"The creator clarifies, “that doesn't mean you should use [text] or [graphs] exclusively. They can be used in conjunction with each other. You can have grep, other types of search, and graph search together. Then, given that the AI has both contexts, it will decide which way to go.”\",[],{\"_37\":4261,\"_23\":313,\"_314\":4262,\"_322\":4267,\"_324\":325},\"55289533b05a\",[4263],{\"_37\":4264,\"_23\":318,\"_319\":4265,\"_174\":4266},\"494b6d9eb54a\",[],\"The creator suggests that the project specifically addresses the dichotomy between graph search and traditional text search/grep. “The first part of the project is how you go from code to graph. The second is an assistant that has a context of a graph, but it also has all the tools that typical coding agents or harnesses have. I also use it alongside my other coding harnesses. For some problems, it can find solutions faster and more accurately.”\",[],{\"_37\":4269,\"_23\":313,\"_314\":4270,\"_322\":4275,\"_324\":408},\"cc2e0c67c8ad\",[4271],{\"_37\":4272,\"_23\":318,\"_319\":4273,\"_174\":4274},\"a7a96359e940\",[],\"Accounting for Variation Across Languages, Frameworks\",[],{\"_37\":4277,\"_23\":313,\"_314\":4278,\"_322\":4301,\"_324\":325},\"755c01bddb6f\",[4279,4283,4288,4292,4297],{\"_37\":4280,\"_23\":318,\"_319\":4281,\"_174\":4282},\"8cfad1aaed2e\",[],\"The creator clarifies that precisely mapping codebases to graphs requires a correct build for each framework and programming language. “Even with an underlying technology like \",{\"_37\":4284,\"_23\":318,\"_319\":4285,\"_174\":4287},\"c4f79c67c17a\",[4286],\"d7c41e85c164\",\"Tree-sitter\",{\"_37\":4289,\"_23\":318,\"_319\":4290,\"_174\":4291},\"2d414b50c68b\",[],\", which parses the \",{\"_37\":4293,\"_23\":318,\"_319\":4294,\"_174\":4296},\"ecb02d6d29a7\",[4295],\"249727404c78\",\"abstract syntax tree\",{\"_37\":4298,\"_23\":318,\"_319\":4299,\"_174\":4300},\"916d54525867\",[],\" and gives you all the syntactic relationships, it's not still enough because there’s so much variety between languages and frameworks.”\",[4302,4304],{\"_37\":4286,\"_23\":49,\"_439\":4303},\"https://tree-sitter.github.io/tree-sitter/\",{\"_37\":4295,\"_23\":49,\"_439\":4305},\"https://en.wikipedia.org/wiki/Abstract_syntax_tree\",{\"_37\":4307,\"_23\":313,\"_314\":4308,\"_322\":4313,\"_324\":325},\"4b92136c2a79\",[4309],{\"_37\":4310,\"_23\":318,\"_319\":4311,\"_174\":4312},\"916fbfa96bd2\",[],\"Avagyan’s approach was to unify the variety within a single graph ontology that worked across multiple languages, from which he designed the project to let users map relationships within a monorepo, even those that contain multiple languages (the project currently supports more than a dozen). With such precise relationships properly mapped, users can use Cypher or another query language (or some sort of tool to map natural language to a query language) to query any part of a codebase as a graph, and compare against semantic search.\",[],{\"_37\":4315,\"_23\":313,\"_314\":4316,\"_322\":4321,\"_324\":325},\"a36ee5a77fd2\",[4317],{\"_37\":4318,\"_23\":318,\"_319\":4319,\"_174\":4320},\"b5e47a104baa\",[],\"“Obviously, this has lots of benefits. If you ask a question about a specific function or a service with semantic search, it looks at the vector similarity and brings you some functions, some of which may have relevant names, but may not be relevant to your actual query. But using graphs, if you can translate your intent into Cypher (with the help of AI), you can basically pinpoint which function you are looking for.” The creator clarifies that the project also supports semantic search as an add-on to support things like code comments, and that in practice, graph parsing and semantic search tend to be complementary.\",[],{\"_37\":4323,\"_23\":313,\"_314\":4324,\"_322\":4329,\"_324\":408},\"6154d54ed4e6\",[4325],{\"_37\":4326,\"_23\":318,\"_319\":4327,\"_174\":4328},\"1641bec46e7a\",[],\"Making Code Graphs Deterministic with AST\",[],{\"_37\":4331,\"_23\":313,\"_314\":4332,\"_322\":4337,\"_324\":325},\"0815be5d9aee\",[4333],{\"_37\":4334,\"_23\":318,\"_319\":4335,\"_174\":4336},\"6ea22a083826\",[],\"As he sought to ensure the project yielded reliable graph representations, Avagyan found that asking LLMs to one-shot graph representations didn’t quite work. “[The LLMs] hallucinated some relationship edges and nodes. So, the project uses Tree-sitter parsers for different languages, which give you the basic relationships you can extract from AST as something of a unified way of using any type of language.”\",[],{\"_37\":4339,\"_23\":313,\"_314\":4340,\"_322\":4354,\"_324\":325},\"4ed1d30f3dc2\",[4341,4345,4350],{\"_37\":4342,\"_23\":318,\"_319\":4343,\"_174\":4344},\"9a4ee9a258b2\",[],\"“Obviously, there are so many considerations when you go into specifics, like finding call relationships. Each language has its own specific intricacies, like how \",{\"_37\":4346,\"_23\":318,\"_319\":4347,\"_174\":4349},\"e56a902ff095\",[4348],\"1b1e54e0707e\",\"polymorphism\",{\"_37\":4351,\"_23\":318,\"_319\":4352,\"_174\":4353},\"9b109a84fb06\",[],\" works.” The creator notes that tracking all the nuances isn’t necessarily covered by existing deterministic open-source tools either. What is even more difficult is the runtime behavior of the code. That's why recent efforts have been placed to trace call relationships apparent at runtime and merge them into the same graph with statically-derived ones.”\",[4355],{\"_37\":4348,\"_23\":49,\"_439\":4356},\"https://en.wikipedia.org/wiki/Polymorphism_(computer_science)\",{\"_37\":4358,\"_23\":313,\"_314\":4359,\"_322\":4373,\"_324\":325},\"9b53b53a2ef7\",[4360,4364,4369],{\"_37\":4361,\"_23\":318,\"_319\":4362,\"_174\":4363},\"84140f1e533e\",[],\"“And from there, we run thousands of test cases to ensure there are no regressions every time a new functionality is added. Every new functionality or bugfix is accompanied by a handful of new tests.” The creator suggests the project continuously runs more than 7,000 unit/integration tests using \",{\"_37\":4365,\"_23\":318,\"_319\":4366,\"_174\":4368},\"a90a0f3ba6d2\",[4367],\"cb676615c060\",\"red/green test-driven development\",{\"_37\":4370,\"_23\":318,\"_319\":4371,\"_174\":4372},\"a3e38bb4ef42\",[],\". Over time, the combination of hand-built code and agent-powered testing and development has proven to be effective. “At the beginning, I spent so much time on architecting [the project], setting the right relationships, that sort of thing, to ensure the foundation was ‘right.’ But when the foundation is ‘right,’ AI is great at doing the rest.”\",[4374],{\"_37\":4367,\"_23\":49,\"_439\":4375},\"https://simonwillison.net/guides/agentic-engineering-patterns/red-green-tdd/\",{\"_37\":4377,\"_23\":313,\"_314\":4378,\"_322\":4383,\"_324\":408},\"e71e561bda30\",[4379],{\"_37\":4380,\"_23\":318,\"_319\":4381,\"_174\":4382},\"8bfe3199c316\",[],\"Bottlenecks and Resource Management\",[],{\"_37\":4385,\"_23\":313,\"_314\":4386,\"_322\":4391,\"_324\":325},\"2e27e83e6ea0\",[4387],{\"_37\":4388,\"_23\":318,\"_319\":4389,\"_174\":4390},\"b4e8f24bb41b\",[],\"The creator suggests that the project is continuing to improve on traditional challenges in graph-based projects, which have historically not always been particularly responsive. “I would say storage has never been a problem. With this project, I spent significant time optimizing the build process of going from codebase to graph. (I actually optimised the build time by more than 5x with a separate repo I built that optimizes user-defined codebase segments against specific metrics like speed within accepted memory constraints. The agent then \\\"autosearches\\\" the best code among competing candidates while running the build against the ground-truthed evaluation cases which you or your agents have collected or constructed yourself.)”\",[],{\"_37\":4393,\"_23\":313,\"_314\":4394,\"_322\":4416,\"_324\":325},\"d8236069c0ba\",[4395,4399,4404,4408,4412],{\"_37\":4396,\"_23\":318,\"_319\":4397,\"_174\":4398},\"37e4c8419854\",[],\"“I would say that a practical use of code-graph-rag is \",{\"_37\":4400,\"_23\":318,\"_319\":4401,\"_174\":4403},\"0f95656bd1a7\",[4402],\"ec14de8b6db9\",\"dead code detection\",{\"_37\":4405,\"_23\":318,\"_319\":4406,\"_174\":4407},\"28ed3f1a3a15\",[],\". If you have vibe coded your repo, or it’s otherwise in a state that you don't know exactly what's happening because there are so many abstractions or code duplications, or that code isn’t reachable. For example, there is a \",{\"_37\":4409,\"_23\":318,\"_319\":4410,\"_174\":4411},\"cba21817906a\",[508],\"code graph\",{\"_37\":4413,\"_23\":318,\"_319\":4414,\"_174\":4415},\"2dd3e29f6f5e\",[],\" command that will check the reachability of all your code. If some are not reachable, they’re likely candidates [for being dead code and potentially getting deleted].”\",[4417],{\"_37\":4402,\"_23\":49,\"_439\":4418},\"https://en.wikipedia.org/wiki/Dead-code_elimination\",{\"_37\":4420,\"_23\":313,\"_314\":4421,\"_322\":4434,\"_324\":325},\"623b0301e0bf\",[4422,4426,4430],{\"_37\":4423,\"_23\":318,\"_319\":4424,\"_174\":4425},\"c9fdbc40e4d6\",[],\"“Otherwise, I don't think storage is a problem. The project uses \",{\"_37\":4427,\"_23\":318,\"_319\":4428,\"_174\":4202},\"34da333815d3\",[4429],\"c1ed526f5769\",{\"_37\":4431,\"_23\":318,\"_319\":4432,\"_174\":4433},\"da89921f17d9\",[],\" as a GraphDB and it's really fast, it runs in memory and you can deploy using Docker containers, and that would be enough to deploy locally, and the project has the building blocks for bigger deployments to the cloud. So far, it seems to scale well (people have even graphed the Linux kernel) and I haven’t seen any real degradation in speed when it comes to querying graphs.”\",[4435],{\"_37\":4429,\"_23\":49,\"_439\":4436},\"https://memgraph.com/\",{\"_37\":4438,\"_23\":313,\"_314\":4439,\"_322\":4452,\"_324\":325},\"824172ebddac\",[4440,4444,4448],{\"_37\":4441,\"_23\":318,\"_319\":4442,\"_174\":4443},\"7cdd05898655\",[],\"Avagyan notes that some users come to the project looking for performance and cost gains as well. “Some people claim that the project is as much as 65% more cost-efficient than traditional agentic coding and search, and I can’t verify those numbers myself. I \",{\"_37\":4445,\"_23\":318,\"_319\":4446,\"_174\":4447},\"36ec44554375\",[508],\"can\",{\"_37\":4449,\"_23\":318,\"_319\":4450,\"_174\":4451},\"e051511f75df\",[],\" say that I’ve observed the project to be very fast, and it seems more fault-tolerant in a way.”\",[],{\"_37\":4454,\"_23\":313,\"_314\":4455,\"_322\":4460,\"_324\":325},\"2cfb4c4a9a81\",[4456],{\"_37\":4457,\"_23\":318,\"_319\":4458,\"_174\":4459},\"b8c18965ccb1\",[],\"“Let’s say the project fails in one Cypher query? Then it can just repeat it, and with that query, you can gain lots of context up front, very fast. And that failure is less costly compared to if AI were to go and perform its own search, realize it was searching in the wrong place, then go to another place...you can imagine how things would accumulate really fast.”\",[],{\"_37\":4462,\"_23\":313,\"_314\":4463,\"_322\":4468,\"_324\":325},\"6a4ddfd50139\",[4464],{\"_37\":4465,\"_23\":318,\"_319\":4466,\"_174\":4467},\"847581b898f9\",[],\"“To be honest, the goal of the project was never cost reduction, but I think it’s a byproduct, since you get the accuracy of a graph and the accuracy of an MCP and agent to invoke the correct tools to query the graph.\",[],{\"_37\":4470,\"_23\":313,\"_314\":4471,\"_322\":4476,\"_324\":408},\"06fc03e6d208\",[4472],{\"_37\":4473,\"_23\":318,\"_319\":4474,\"_174\":4475},\"0743e3b412d6\",[],\"Real-World Use Cases for Graphing a Codebase\",[],{\"_37\":4478,\"_23\":313,\"_314\":4479,\"_322\":4484,\"_324\":325},\"34664692f09c\",[4480],{\"_37\":4481,\"_23\":318,\"_319\":4482,\"_174\":4483},\"2b1eaf954306\",[],\"The creator calls out onboarding new engineers as a key use case for the project. “People tell me that using the project has significantly reduced onboarding times because it can gather information about different aspects of a monorepo codebase and give that directly to new junior engineers.”\",[],{\"_37\":4486,\"_23\":313,\"_314\":4487,\"_322\":4492,\"_324\":325},\"a9bdfd671d83\",[4488],{\"_37\":4489,\"_23\":318,\"_319\":4490,\"_174\":4491},\"b0569066d73c\",[],\"“Since we recently added FLOWS_TO, READS_FROM and WRITES_TO relationships in a graph, it’s now easy to analyze the relationships between microservices in your codebase as well. (I’ve actually been approached by security companies about using the project for security detection use cases across large codebases, which is a natural fit as tracing how data moves through a system is exactly what those relationships capture.)”\",[],{\"_37\":4494,\"_23\":313,\"_314\":4495,\"_322\":4500,\"_324\":325},\"e96ed7f2d62e\",[4496],{\"_37\":4497,\"_23\":318,\"_319\":4498,\"_174\":4499},\"e4b2c56f631d\",[],\"So you can see why companies are adopting the project as part of their onboarding tools. You could use it as an MCP server in Claude Code or any other harness or just use Cypher queries to search your codebase before even starting any agentic exploration. This approach is quick and gives you a really good grasp of a codebase, without getting sidetracked to other areas that may not be as important.”\",[],{\"_37\":4502,\"_23\":313,\"_314\":4503,\"_322\":4526,\"_324\":325},\"b003dd295bbf\",[4504,4508,4513,4517,4522],{\"_37\":4505,\"_23\":318,\"_319\":4506,\"_174\":4507},\"2a8e4e8ba899\",[],\"Avagyan notes growing interest among teams managing heavyweight ERP systems with some flavor of Python, as well as ongoing requests around app development. “\",{\"_37\":4509,\"_23\":318,\"_319\":4510,\"_174\":4512},\"fc85ec3d18b6\",[4511],\"fc884ca87ef9\",\"Dart\",{\"_37\":4514,\"_23\":318,\"_319\":4515,\"_174\":4516},\"cee0616ee44a\",[],\" is already supported and includes \",{\"_37\":4518,\"_23\":318,\"_319\":4519,\"_174\":4521},\"ab660c7a1ffd\",[4520],\"eb5b74dd23fd\",\"Flutter\",{\"_37\":4523,\"_23\":318,\"_319\":4524,\"_174\":4525},\"b630193673be\",[],\" widget structure.”\",[4527,4529],{\"_37\":4511,\"_23\":49,\"_439\":4528},\"https://dart.dev/overview\",{\"_37\":4520,\"_23\":49,\"_439\":4530},\"https://flutter.dev/\",{\"_37\":4532,\"_23\":313,\"_314\":4533,\"_322\":4538,\"_324\":408},\"5a3ea8acd83b\",[4534],{\"_37\":4535,\"_23\":318,\"_319\":4536,\"_174\":4537},\"1cd77c6dd0f6\",[],\"Why Aren’t Enterprises Already Graphing Their Codebases?\",[],{\"_37\":4540,\"_23\":313,\"_314\":4541,\"_322\":4546,\"_324\":325},\"7cbaec07cf9b\",[4542],{\"_37\":4543,\"_23\":318,\"_319\":4544,\"_174\":4545},\"52c6bc34d2b2\",[],\"The creator suggests that graphing codebases is a useful approach for any organization, but the timing might not be quite right for large-scale enterprises that are slow to embrace change. “I believe many companies, including major AI vendors like Cursor or Anthropic, have been experimenting with graph-based approaches, if not for codebase mapping, then for use cases like memory management.”\",[],{\"_37\":4548,\"_23\":313,\"_314\":4549,\"_322\":4554,\"_324\":325},\"a89686ddcbd7\",[4550],{\"_37\":4551,\"_23\":318,\"_319\":4552,\"_174\":4553},\"eb00873aac4a\",[],\"Avagyan notes that those companies may be hitting a stumbling block. “I think one problem companies may be having is accuracy.”\",[],{\"_37\":4556,\"_23\":313,\"_314\":4557,\"_322\":4562,\"_324\":325},\"f2f91f55c9c2\",[4558],{\"_37\":4559,\"_23\":318,\"_319\":4560,\"_174\":4561},\"5556895009c6\",[],\"“You can get really far with one language, maybe with a toy example using a tool like Tree-sitter. But jobs that require you to unify graphs across different languages, and build usable and accurate relationships over top of them, and making sure it can work at scale, that’s different. For example, a monorepo is a living creature that changes every minute: People commit and delete code and so on. So your graph needs to keep up with that. It needs to recalculate [those relationships] with every change, every drop, every addition. The project has a graph updater designed specifically to make this real-time syncing possible.”\",[],{\"_37\":4564,\"_23\":313,\"_314\":4565,\"_322\":4570,\"_324\":325},\"526da0bc67f6\",[4566],{\"_37\":4567,\"_23\":318,\"_319\":4568,\"_174\":4569},\"b5c434461aac\",[],\"The creator suggests that while he spent the better part of a year building a system that could dynamically map and remap graph representations of codebases without significant latency, not every org is on the same page. “Even doing something like this with AI agents will take some time, and will cost internal resources. And people don’t always have patience for these things and may want to see a one-shot solution instead. But I think the approach will get more momentum and I think many companies will adopt it eventually.”\",[],{\"_37\":4572,\"_23\":313,\"_314\":4573,\"_322\":4578,\"_324\":408},\"05e8e1c9fe0a\",[4574],{\"_37\":4575,\"_23\":318,\"_319\":4576,\"_174\":4577},\"76872200c049\",[],\"Learnings from Building\",[],{\"_37\":4580,\"_23\":313,\"_314\":4581,\"_322\":4586,\"_324\":325},\"8c4981547047\",[4582],{\"_37\":4583,\"_23\":318,\"_319\":4584,\"_174\":4585},\"58335d9df518\",[],\"When asked what his most important learnings from the project are, Avagyan offers a few suggestions. “One, which I think is most important, is to just build something for yourself that you would like to have. That's pretty much how this project was born. Because I was dealing with [the challenges of managing a] monorepo, and at first, I just wanted to visualize it and understand its complexity. I looked for solutions to see if there was a good product for me. I didn't find one, so I decided I would build it myself, and after releasing it publicly, it turned to be useful for other people.”\",[],{\"_37\":4588,\"_23\":313,\"_314\":4589,\"_322\":4620,\"_324\":325},\"81d7413fc3b2\",[4590,4594,4599,4603,4608,4611,4616],{\"_37\":4591,\"_23\":318,\"_319\":4592,\"_174\":4593},\"321ac6bc0e40\",[],\"“Second, I would recommend building things from the perspective of thinking about what you do every day and how you can accelerate what you do and be more performant with your tools.” The creator notes another open-source project of his, \",{\"_37\":4595,\"_23\":318,\"_319\":4596,\"_174\":4598},\"9f8ed6fdab01\",[4597],\"f040745fb71d\",\"Croft\",{\"_37\":4600,\"_23\":318,\"_319\":4601,\"_174\":4602},\"630e70cbbe50\",[],\", as an example. “I’m a VS Code user, but I enjoy terminal as well, so I found myself going back and forth between VS Code, \",{\"_37\":4604,\"_23\":318,\"_319\":4605,\"_174\":4607},\"94d7ac6fa32a\",[4606],\"3f971d287b17\",\"Neovim\",{\"_37\":4609,\"_23\":318,\"_319\":4610,\"_174\":679},\"e76bcfb5a8f3\",[],{\"_37\":4612,\"_23\":318,\"_319\":4613,\"_174\":4615},\"93a91dcf4857\",[4614],\"2f238ed440f7\",\"Zed\",{\"_37\":4617,\"_23\":318,\"_319\":4618,\"_174\":4619},\"1bb94c846e83\",[],\" and other alternatives. I found myself thinking there should be a better way, so I built a VS Code clone in the terminal with Rust to make it super-fast. I’ve been using it every day myself and I would recommend others try it out.”\",[4621,4623,4625],{\"_37\":4597,\"_23\":49,\"_439\":4622},\"https://github.com/vitali87/croft\",{\"_37\":4606,\"_23\":49,\"_439\":4624},\"https://neovim.io/\",{\"_37\":4614,\"_23\":49,\"_439\":4626},\"https://zed.dev/\",{\"_37\":4628,\"_23\":313,\"_314\":4629,\"_322\":4652,\"_324\":325},\"6869a525750a\",[4630,4634,4639,4643,4648],{\"_37\":4631,\"_23\":318,\"_319\":4632,\"_174\":4633},\"30637c109bb6\",[],\"“Finally, I’d recommend finding a distribution channel for your work. Even the greatest product on earth won’t necessarily be noticed if you just put it on GitHub and leave it there. You need to promote it. I was fortunate enough to have worked with the founder of \",{\"_37\":4635,\"_23\":318,\"_319\":4636,\"_174\":4638},\"6424a1b96d65\",[4637],\"48182a105088\",\"Daily Dose of Data Science\",{\"_37\":4640,\"_23\":318,\"_319\":4641,\"_174\":4642},\"3954a9df1f48\",[],\" when I was working on my own startup back in 2023. I got coverage for my product in this popular newsletter, which might have been the trigger that helped it ‘go viral’ on GitHub the first time. That said, I have also tried many other channels like Hacker News, LinkedIn and X posts, giving talks (including this session \",{\"_37\":4644,\"_23\":318,\"_319\":4645,\"_174\":4647},\"6f0d56991c70\",[4646],\"41760ee29615\",\"with Memgraph\",{\"_37\":4649,\"_23\":318,\"_319\":4650,\"_174\":4651},\"dd3d4a61f74f\",[],\"), but eventually, the project started trending again on GitHub organically. I guess it might have to do with the ‘right solution at the right time’ phenomenon.”\",[4653,4655],{\"_37\":4637,\"_23\":49,\"_439\":4654},\"https://blog.dailydoseofds.com/\",{\"_37\":4646,\"_23\":49,\"_439\":4656},\"https://www.youtube.com/watch?v=O2jUTq6nCEY\u0026t=2s\",{\"_37\":4658,\"_23\":313,\"_314\":4659,\"_322\":4664,\"_324\":408},\"1d6c6865362e\",[4660],{\"_37\":4661,\"_23\":318,\"_319\":4662,\"_174\":4663},\"2586e6d9e8b9\",[],\"The Future of Codebases: Built and Managed by AI?\",[],{\"_37\":4666,\"_23\":313,\"_314\":4667,\"_322\":4672,\"_324\":325},\"ab8bfa185ba6\",[4668],{\"_37\":4669,\"_23\":318,\"_319\":4670,\"_174\":4671},\"067bce70e73c\",[],\"From his experience building the project, Avagyan advises anyone working in a technical capacity to learn how to orchestrate. “Software engineering is changing faster than people think. The ability to orchestrate things will be important no matter what. By that, I mean knowing many different technologies: Their strengths and weaknesses, the scalability of each thing, and when to use what.”\",[],{\"_37\":4674,\"_23\":313,\"_314\":4675,\"_322\":4688,\"_324\":325},\"afb3c879f969\",[4676,4680,4684],{\"_37\":4677,\"_23\":318,\"_319\":4678,\"_174\":4679},\"49750bd53c8c\",[],\"“For example, in order to deploy applications at scale, you might need a Kubernetes cluster. You shouldn’t have to be a highly skilled DevOps engineer to know that as AI can help you there, but you would need to know \",{\"_37\":4681,\"_23\":318,\"_319\":4682,\"_174\":4683},\"7f5989d240bd\",[508],\"when\",{\"_37\":4685,\"_23\":318,\"_319\":4686,\"_174\":4687},\"56e7f92d1742\",[],\" to use it, and what the pros and cons are. That will help you to have this overarching, holistic view of what it's like to build a highly scalable software engineering system.”\",[],{\"_37\":4690,\"_23\":313,\"_314\":4691,\"_322\":4696,\"_324\":325},\"dc75d7ad8097\",[4692],{\"_37\":4693,\"_23\":318,\"_319\":4694,\"_174\":4695},\"c6a9980c57d2\",[],\"“You need to understand how the distributed systems work, and the code-graph-rag project fits really well here to help you ask architectural questions and locate potential bottlenecks, like dead code or other opportunities for optimization.”\",[],{\"_37\":4698,\"_23\":313,\"_314\":4699,\"_322\":4712,\"_324\":325},\"628421de9da8\",[4700,4704,4708],{\"_37\":4701,\"_23\":318,\"_319\":4702,\"_174\":4703},\"0cc52c592750\",[],\"“It’s about having this type of ‘systems thinking,’ an understanding of system design architecture and the underlying principles of the technologies. Not just knowing \",{\"_37\":4705,\"_23\":318,\"_319\":4706,\"_174\":4707},\"d4b8c4a82cf1\",[508],\"what Redis is\",{\"_37\":4709,\"_23\":318,\"_319\":4710,\"_174\":4711},\"ad91e713dbfb\",[],\", but also understanding what caching is and why it’s important. You’re going to need to apply that understanding when you think your system actually needs caching!”\",[],{\"_37\":4714,\"_23\":313,\"_314\":4715,\"_322\":4720,\"_324\":325},\"8eaf5dda166e\",[4716],{\"_37\":4717,\"_23\":318,\"_319\":4718,\"_174\":4719},\"7fccb285f275\",[],\"“So I think orchestration will be important for quite some time, because AI is still not at the level of fully, autonomously orchestrating large-scale systems that will work smoothly and without any problems. We are not there yet.”\",[],\"https://cdn.sanity.io/images/50q6fr1p/production/28801f05c737cc57c3c04d467a0a54838fc70ad0-1200x630.png\",\"9d8377ac-25e6-47b7-8e40-22ee9a7cb43e\",\"2026-08-27T16:38:00.000Z\",\"understanding-code-by-treating-it-as-a-graph\",\"Is the Key to Understanding Code Treating it as a Graph?\",{\"_398\":4727,\"_2820\":4110,\"_149\":4111,\"_3617\":4112,\"_912\":4113,\"_366\":4114,\"_32\":4115,\"_3621\":278},[4728,4733,4738,4761,4766,4768,4773,4778,4793,4798,4803,4808,4813,4818,4823,4840,4847,4854,4861,4868,4873,4878,4883,4888,4893,4898,4903,4918,4923,4928,4933,4938,4943,4948,4953,4958,4963,4968],{\"_37\":3687,\"_23\":313,\"_314\":4729,\"_322\":4732,\"_324\":408},[4730],{\"_37\":3690,\"_23\":318,\"_319\":4731,\"_174\":3692},[],[],{\"_37\":3695,\"_23\":313,\"_314\":4734,\"_322\":4737,\"_324\":325},[4735],{\"_37\":3698,\"_23\":318,\"_319\":4736,\"_174\":3700},[],[],{\"_37\":3703,\"_23\":313,\"_314\":4739,\"_322\":4756,\"_324\":325},[4740,4742,4744,4746,4748,4750,4752,4754],{\"_37\":3706,\"_23\":318,\"_319\":4741,\"_174\":3709},[3708],{\"_37\":3711,\"_23\":318,\"_319\":4743,\"_174\":3713},[],{\"_37\":3715,\"_23\":318,\"_319\":4745,\"_174\":3718},[3717],{\"_37\":3720,\"_23\":318,\"_319\":4747,\"_174\":846},[],{\"_37\":3723,\"_23\":318,\"_319\":4749,\"_174\":3726},[3725],{\"_37\":3728,\"_23\":318,\"_319\":4751,\"_174\":3730},[],{\"_37\":3732,\"_23\":318,\"_319\":4753,\"_174\":3735},[3734],{\"_37\":3737,\"_23\":318,\"_319\":4755,\"_174\":3739},[],[4757,4758,4759,4760],{\"_37\":3708,\"_23\":49,\"_439\":3742},{\"_37\":3717,\"_23\":49,\"_439\":3744},{\"_37\":3725,\"_23\":49,\"_439\":3746},{\"_37\":3734,\"_23\":49,\"_439\":3748},{\"_37\":3750,\"_23\":313,\"_314\":4762,\"_322\":4765,\"_324\":325},[4763],{\"_37\":3753,\"_23\":318,\"_319\":4764,\"_174\":3755},[],[],{\"_37\":3758,\"_23\":149,\"_114\":4767},{\"_116\":3760,\"_23\":118},{\"_37\":3762,\"_23\":313,\"_314\":4769,\"_322\":4772,\"_324\":325},[4770],{\"_37\":3765,\"_23\":318,\"_319\":4771,\"_174\":3767},[508],[],{\"_37\":3770,\"_23\":313,\"_314\":4774,\"_322\":4777,\"_324\":408},[4775],{\"_37\":3773,\"_23\":318,\"_319\":4776,\"_174\":3775},[],[],{\"_37\":3778,\"_23\":313,\"_314\":4779,\"_322\":4790,\"_324\":325},[4780,4782,4784,4786,4788],{\"_37\":3781,\"_23\":318,\"_319\":4781,\"_174\":3783},[],{\"_37\":3785,\"_23\":318,\"_319\":4783,\"_174\":3788},[3787],{\"_37\":3790,\"_23\":318,\"_319\":4785,\"_174\":846},[],{\"_37\":3793,\"_23\":318,\"_319\":4787,\"_174\":3796},[3795],{\"_37\":3798,\"_23\":318,\"_319\":4789,\"_174\":3800},[],[4791,4792],{\"_37\":3787,\"_23\":49,\"_439\":3803},{\"_37\":3795,\"_23\":49,\"_439\":3805},{\"_37\":3807,\"_23\":313,\"_314\":4794,\"_322\":4797,\"_324\":325},[4795],{\"_37\":3810,\"_23\":318,\"_319\":4796,\"_174\":3812},[],[],{\"_37\":3815,\"_23\":313,\"_314\":4799,\"_322\":4802,\"_324\":325},[4800],{\"_37\":3818,\"_23\":318,\"_319\":4801,\"_174\":3820},[],[],{\"_37\":3823,\"_23\":313,\"_314\":4804,\"_322\":4807,\"_324\":325},[4805],{\"_37\":3826,\"_23\":318,\"_319\":4806,\"_174\":3828},[],[],{\"_37\":3831,\"_23\":313,\"_314\":4809,\"_322\":4812,\"_324\":408},[4810],{\"_37\":3834,\"_23\":318,\"_319\":4811,\"_174\":3836},[],[],{\"_37\":3839,\"_23\":313,\"_314\":4814,\"_322\":4817,\"_324\":325},[4815],{\"_37\":3842,\"_23\":318,\"_319\":4816,\"_174\":3844},[],[],{\"_37\":3847,\"_23\":313,\"_314\":4819,\"_322\":4822,\"_324\":325},[4820],{\"_37\":3850,\"_23\":318,\"_319\":4821,\"_174\":3852},[],[],{\"_37\":3855,\"_23\":313,\"_314\":4824,\"_2171\":1957,\"_2172\":2173,\"_322\":4837,\"_324\":325},[4825,4827,4829,4831,4833,4835],{\"_37\":3858,\"_23\":318,\"_319\":4826,\"_174\":3860},[2165],{\"_37\":3862,\"_23\":318,\"_319\":4828,\"_174\":3864},[],{\"_37\":3866,\"_23\":318,\"_319\":4830,\"_174\":3869},[3868],{\"_37\":3871,\"_23\":318,\"_319\":4832,\"_174\":3873},[],{\"_37\":3875,\"_23\":318,\"_319\":4834,\"_174\":3878},[3877],{\"_37\":3880,\"_23\":318,\"_319\":4836,\"_174\":3882},[],[4838,4839],{\"_37\":3868,\"_23\":49,\"_439\":3885},{\"_37\":3877,\"_23\":49,\"_439\":3887},{\"_37\":3889,\"_23\":313,\"_314\":4841,\"_2171\":1957,\"_2172\":2173,\"_322\":4846,\"_324\":325},[4842,4844],{\"_37\":3892,\"_23\":318,\"_319\":4843,\"_174\":3894},[2165],{\"_37\":3896,\"_23\":318,\"_319\":4845,\"_174\":3898},[],[],{\"_37\":3901,\"_23\":313,\"_314\":4848,\"_2171\":1957,\"_2172\":2173,\"_322\":4853,\"_324\":325},[4849,4851],{\"_37\":3904,\"_23\":318,\"_319\":4850,\"_174\":3906},[2165],{\"_37\":3908,\"_23\":318,\"_319\":4852,\"_174\":3910},[],[],{\"_37\":3913,\"_23\":313,\"_314\":4855,\"_2171\":1957,\"_2172\":2173,\"_322\":4860,\"_324\":325},[4856,4858],{\"_37\":3916,\"_23\":318,\"_319\":4857,\"_174\":3918},[2165],{\"_37\":3920,\"_23\":318,\"_319\":4859,\"_174\":3922},[],[],{\"_37\":3925,\"_23\":313,\"_314\":4862,\"_2171\":1957,\"_2172\":2173,\"_322\":4867,\"_324\":325},[4863,4865],{\"_37\":3928,\"_23\":318,\"_319\":4864,\"_174\":3930},[2165],{\"_37\":3932,\"_23\":318,\"_319\":4866,\"_174\":3934},[],[],{\"_37\":3937,\"_23\":313,\"_314\":4869,\"_322\":4872,\"_324\":325},[4870],{\"_37\":3940,\"_23\":318,\"_319\":4871,\"_174\":3942},[],[],{\"_37\":3945,\"_23\":313,\"_314\":4874,\"_322\":4877,\"_324\":325},[4875],{\"_37\":3948,\"_23\":318,\"_319\":4876,\"_174\":3950},[],[],{\"_37\":3953,\"_23\":313,\"_314\":4879,\"_322\":4882,\"_324\":325},[4880],{\"_37\":3956,\"_23\":318,\"_319\":4881,\"_174\":3958},[],[],{\"_37\":3961,\"_23\":313,\"_314\":4884,\"_322\":4887,\"_324\":408},[4885],{\"_37\":3964,\"_23\":318,\"_319\":4886,\"_174\":3966},[],[],{\"_37\":3969,\"_23\":313,\"_314\":4889,\"_322\":4892,\"_324\":325},[4890],{\"_37\":3972,\"_23\":318,\"_319\":4891,\"_174\":3974},[],[],{\"_37\":3977,\"_23\":313,\"_314\":4894,\"_322\":4897,\"_324\":325},[4895],{\"_37\":3980,\"_23\":318,\"_319\":4896,\"_174\":3982},[],[],{\"_37\":3985,\"_23\":313,\"_314\":4899,\"_322\":4902,\"_324\":325},[4900],{\"_37\":3988,\"_23\":318,\"_319\":4901,\"_174\":3990},[],[],{\"_37\":3993,\"_23\":313,\"_314\":4904,\"_322\":4915,\"_324\":325},[4905,4907,4909,4911,4913],{\"_37\":3996,\"_23\":318,\"_319\":4906,\"_174\":3998},[],{\"_37\":4000,\"_23\":318,\"_319\":4908,\"_174\":4003},[4002],{\"_37\":4005,\"_23\":318,\"_319\":4910,\"_174\":4007},[],{\"_37\":4009,\"_23\":318,\"_319\":4912,\"_174\":4012},[4011],{\"_37\":4014,\"_23\":318,\"_319\":4914,\"_174\":4016},[],[4916,4917],{\"_37\":4002,\"_23\":49,\"_439\":4019},{\"_37\":4011,\"_23\":49,\"_439\":4021},{\"_37\":4023,\"_23\":313,\"_314\":4919,\"_322\":4922,\"_324\":325},[4920],{\"_37\":4026,\"_23\":318,\"_319\":4921,\"_174\":4028},[],[],{\"_37\":4031,\"_23\":313,\"_314\":4924,\"_322\":4927,\"_324\":325},[4925],{\"_37\":4034,\"_23\":318,\"_319\":4926,\"_174\":4036},[],[],{\"_37\":4039,\"_23\":313,\"_314\":4929,\"_322\":4932,\"_324\":325},[4930],{\"_37\":4042,\"_23\":318,\"_319\":4931,\"_174\":4044},[],[],{\"_37\":4047,\"_23\":313,\"_314\":4934,\"_322\":4937,\"_324\":325},[4935],{\"_37\":4050,\"_23\":318,\"_319\":4936,\"_174\":4052},[],[],{\"_37\":4055,\"_23\":313,\"_314\":4939,\"_322\":4942,\"_324\":408},[4940],{\"_37\":4058,\"_23\":318,\"_319\":4941,\"_174\":4060},[],[],{\"_37\":4063,\"_23\":313,\"_314\":4944,\"_322\":4947,\"_324\":325},[4945],{\"_37\":4066,\"_23\":318,\"_319\":4946,\"_174\":4068},[],[],{\"_37\":4071,\"_23\":313,\"_314\":4949,\"_322\":4952,\"_324\":325},[4950],{\"_37\":4074,\"_23\":318,\"_319\":4951,\"_174\":4076},[],[],{\"_37\":4079,\"_23\":313,\"_314\":4954,\"_322\":4957,\"_324\":325},[4955],{\"_37\":4082,\"_23\":318,\"_319\":4956,\"_174\":4084},[],[],{\"_37\":4087,\"_23\":313,\"_314\":4959,\"_322\":4962,\"_324\":325},[4960],{\"_37\":4090,\"_23\":318,\"_319\":4961,\"_174\":4092},[],[],{\"_37\":4095,\"_23\":313,\"_314\":4964,\"_322\":4967,\"_324\":325},[4965],{\"_37\":4098,\"_23\":318,\"_319\":4966,\"_174\":4100},[],[],{\"_37\":4103,\"_23\":313,\"_314\":4969,\"_322\":4972,\"_324\":325},[4970],{\"_37\":4106,\"_23\":318,\"_319\":4971,\"_174\":4108},[],[],{\"_398\":4974,\"_2820\":5694,\"_149\":5695,\"_3617\":5696,\"_912\":5697,\"_366\":5698,\"_32\":5699,\"_3621\":278},[4975,4983,5013,5021,5068,5072,5087,5095,5114,5133,5148,5156,5186,5194,5202,5210,5229,5248,5264,5272,5280,5288,5296,5304,5312,5331,5366,5374,5398,5417,5425,5433,5449,5457,5465,5480,5488,5496,5504,5523,5547,5573,5581,5608,5616,5624,5632,5640,5648,5667,5686],{\"_37\":4976,\"_23\":313,\"_314\":4977,\"_322\":4982,\"_324\":408},\"7a24ac14d07e\",[4978],{\"_37\":4979,\"_23\":318,\"_319\":4980,\"_174\":4981},\"48d3443f20a8\",[],\"How Can Agents Be the Future if They’re not Reliable?\",[],{\"_37\":4984,\"_23\":313,\"_314\":4985,\"_322\":5008,\"_324\":325},\"ff2d4b14fb40\",[4986,4990,4995,4999,5004],{\"_37\":4987,\"_23\":318,\"_319\":4988,\"_174\":4989},\"4b0676b5544d\",[],\"Some reports suggest that as many as \",{\"_37\":4991,\"_23\":318,\"_319\":4992,\"_174\":4994},\"7633cd943027\",[4993],\"72307ba313ce\",\"25%\",{\"_37\":4996,\"_23\":318,\"_319\":4997,\"_174\":4998},\"103406d76f91\",[],\" of enterprises have adopted agentic workflows. Isn’t this the era of AI agents? Why isn’t the number higher? Aside from predictable enterprise-level concerns (\",{\"_37\":5000,\"_23\":318,\"_319\":5001,\"_174\":5003},\"01325be45d21\",[5002],\"c2f310fcffb7\",\"49%\",{\"_37\":5005,\"_23\":318,\"_319\":5006,\"_174\":5007},\"c145585f2175\",[],\" of enterprises are reportedly holding off due to security concerns, for instance), agents haven’t proven to be reliable workhorses at scale, particularly for long-horizon tasks. Not yet, anyway.\",[5009,5011],{\"_37\":4993,\"_23\":49,\"_439\":5010},\"https://snyk.io/blog/agentic-development-lifecycle/\",{\"_37\":5002,\"_23\":49,\"_439\":5012},\"https://www.forrester.com/blogs/the-state-of-agentic-ai-in-2026-companies-are-chasing-few-are-catching/\",{\"_37\":5014,\"_23\":313,\"_314\":5015,\"_322\":5020,\"_324\":325},\"72359702785e\",[5016],{\"_37\":5017,\"_23\":318,\"_319\":5018,\"_174\":5019},\"c805aa2c35cb\",[],\"Agents still can’t “remember” their learnings from previous successful runs, still hallucinate when their context window becomes full, and still have a number of other issues. Wasn’t the perfect agentic future supposed to feature technical teams commanding fleets of infallible, well-trained agents to do their bidding?\",[],{\"_37\":5022,\"_23\":313,\"_314\":5023,\"_322\":5061,\"_324\":325},\"24ebf08aefeb\",[5024,5028,5033,5037,5042,5045,5048,5052,5057],{\"_37\":5025,\"_23\":318,\"_319\":5026,\"_174\":5027},\"0008706ad2a9\",[],\"Open-source creator \",{\"_37\":5029,\"_23\":318,\"_319\":5030,\"_174\":5032},\"bfecdcdc2d8e\",[5031],\"f4b6e64c3af0\",\"Mike Hostetler\",{\"_37\":5034,\"_23\":318,\"_319\":5035,\"_174\":5036},\"c55ab4189631\",[],\" believes such a future is possible if systems combine strong guardrails with fault-tolerant infrastructure designed for large-scale deployments. Here, he explains why he built his project, \",{\"_37\":5038,\"_23\":318,\"_319\":5039,\"_174\":5041},\"1b8ae31d51e4\",[5040],\"b47c6c24e0ab\",\"Jido\",{\"_37\":5043,\"_23\":318,\"_319\":5044,\"_174\":679},\"4ac8dc73cc44\",[],{\"_37\":5046,\"_23\":318,\"_319\":5047,\"_174\":1534},\"2f226b078804\",[508],{\"_37\":5049,\"_23\":318,\"_319\":5050,\"_174\":5051},\"7c9c7486bda0\",[],\" using Python, the de facto programming language for AI, but using \",{\"_37\":5053,\"_23\":318,\"_319\":5054,\"_174\":5056},\"be12332c52e1\",[5055],\"f60c98fe6c05\",\"Elixir\",{\"_37\":5058,\"_23\":318,\"_319\":5059,\"_174\":5060},\"1edf6210f9e9\",[],\", which was inherently designed to support maintainability and massive concurrency.\",[5062,5064,5066],{\"_37\":5031,\"_23\":49,\"_439\":5063},\"https://www.linkedin.com/in/mikehostetler/\",{\"_37\":5040,\"_23\":49,\"_439\":5065},\"https://github.com/agentjido/jido\",{\"_37\":5055,\"_23\":49,\"_439\":5067},\"https://elixir-lang.org/\",{\"_37\":5069,\"_23\":149,\"_114\":5070},\"4bd56dd08d69\",{\"_116\":5071,\"_23\":118},\"image-734e0114ac063680123b3471dd5f4f588375e8fa-1280x591-jpg\",{\"_37\":5073,\"_23\":313,\"_314\":5074,\"_322\":5084,\"_324\":325},\"b03c891ae9f1\",[5075,5079],{\"_37\":5076,\"_23\":318,\"_319\":5077,\"_174\":5078},\"0c2b5b8ee499\",[508],\"Mike Hostetler discusses how AI is changing the business of building software on the Software Without Borders podcast. Image courtesy \",{\"_37\":5080,\"_23\":318,\"_319\":5081,\"_174\":5083},\"a2615c1217b2\",[5082,508],\"ea2317db32a8\",\"Software \\nWithout Borders\",[5085],{\"_37\":5082,\"_23\":49,\"_439\":5086},\"https://www.youtube.com/watch?v=zVYqx_FDTPU\",{\"_37\":5088,\"_23\":313,\"_314\":5089,\"_322\":5094,\"_324\":408},\"378419cb6538\",[5090],{\"_37\":5091,\"_23\":318,\"_319\":5092,\"_174\":5093},\"b95d6349db96\",[],\"Resolving Nondeterministic LLMs and Deterministic Algorithms\",[],{\"_37\":5096,\"_23\":313,\"_314\":5097,\"_322\":5111,\"_324\":325},\"87b25977462d\",[5098,5102,5107],{\"_37\":5099,\"_23\":318,\"_319\":5100,\"_174\":5101},\"2cb97e117942\",[],\"Hostetler, a veteran developer and founder, got his start working on the \",{\"_37\":5103,\"_23\":318,\"_319\":5104,\"_174\":5106},\"5df850edc9eb\",[5105],\"e7748844a8dd\",\"jQuery\",{\"_37\":5108,\"_23\":318,\"_319\":5109,\"_174\":5110},\"e6f957aa936c\",[],\" ecosystem, from which he built a career, a startup, and a lifelong curiosity about technology. He explains that his introduction to Elixir came from a tour of duty working in distributed systems, which inspired him to think through the challenge of being able to run multiple agents with real-world reliability. “I envisioned a future of ‘10,000 agents per human.’”\",[5112],{\"_37\":5105,\"_23\":49,\"_439\":5113},\"https://jquery.com\",{\"_37\":5115,\"_23\":313,\"_314\":5116,\"_322\":5130,\"_324\":325},\"0609f26003a6\",[5117,5121,5126],{\"_37\":5118,\"_23\":318,\"_319\":5119,\"_174\":5120},\"11d3ae480877\",[],\"The creator notes that his original vision was running 10,000 agents on a single \",{\"_37\":5122,\"_23\":318,\"_319\":5123,\"_174\":5125},\"bd8a2854d8b0\",[5124],\"f82f2e7f077d\",\"Raspberry Pi\",{\"_37\":5127,\"_23\":318,\"_319\":5128,\"_174\":5129},\"3efb34af94bf\",[],\". In other words, in a resource-constrained environment with no direct access to LLMs. Models were getting better, but capital expenditure from foundation labs continued to skyrocket.\",[5131],{\"_37\":5124,\"_23\":49,\"_439\":5132},\"https://www.raspberrypi.com/\",{\"_37\":5134,\"_23\":313,\"_314\":5135,\"_322\":5147,\"_324\":325},\"c7d5705ac245\",[5136,5140,5144],{\"_37\":5137,\"_23\":318,\"_319\":5138,\"_174\":5139},\"22d5ed497f6c\",[],\"Hostetler suggests his goal was to make the best of both worlds: Capitalize on the improving performance of state-of-the-art models, combined with performant infrastructure that wouldn’t break the bank. “The intersection of these two ideas is: Agents needed to be easy to build, easy to deploy, easy to orchestrate and coordinate. And then, 20% of your agents could use an LLM as the thinking step. But it needed to be just as easy to build 80% of your agents on what I call \",{\"_37\":5141,\"_23\":318,\"_319\":5142,\"_174\":5143},\"ae3aaa793b3f\",[508],\"classical AI algorithms\",{\"_37\":5145,\"_23\":318,\"_319\":5146,\"_174\":1518},\"c8c502a31d12\",[],[],{\"_37\":5149,\"_23\":313,\"_314\":5150,\"_322\":5155,\"_324\":325},\"5e457368ab04\",[5151],{\"_37\":5152,\"_23\":318,\"_319\":5153,\"_174\":5154},\"ed9ee37193c9\",[],\"The builder clarifies that ‘classic algorithms’ also include logic that predates LLMs: “Everything from a finite state machine to a behavior tree to hierarchical task networks. Video game AI as well as LLM AI. And those could work in concert and then be easily and inexpensively deployable for a user, which gets into the real pragmatic piece.”\",[],{\"_37\":5157,\"_23\":313,\"_314\":5158,\"_322\":5181,\"_324\":325},\"c843c70bed8d\",[5159,5163,5168,5172,5177],{\"_37\":5160,\"_23\":318,\"_319\":5161,\"_174\":5162},\"9425a3cff2d7\",[],\"Hostetler clarifies that despite Jido’s explicit “AI optional” clause, his approach to AI is less skeptical and more pragmatic. “I love the nondeterminism of LLMs. I love pushing those boundaries. Just by happenstance, having put my own open-source project out there, I got to know \",{\"_37\":5164,\"_23\":318,\"_319\":5165,\"_174\":5167},\"9a152e0af025\",[5166],\"c4b6a4a077f8\",\"Geoff Huntley\",{\"_37\":5169,\"_23\":318,\"_319\":5170,\"_174\":5171},\"f8941b7433c0\",[],\" and tried \",{\"_37\":5173,\"_23\":318,\"_319\":5174,\"_174\":5176},\"b9826d2e81c5\",[5175],\"cfe50a2150c7\",\"the Ralph Wiggum loop\",{\"_37\":5178,\"_23\":318,\"_319\":5179,\"_174\":5180},\"64bca5e784fe\",[],\" before it was publicly available. Loved it, and had Jido running loops.”\",[5182,5184],{\"_37\":5166,\"_23\":49,\"_439\":5183},\"https://ghuntley.com\",{\"_37\":5175,\"_23\":49,\"_439\":5185},\"https://ghuntley.com/loop/\",{\"_37\":5187,\"_23\":313,\"_314\":5188,\"_322\":5193,\"_324\":325},\"103cd73a04dd\",[5189],{\"_37\":5190,\"_23\":318,\"_319\":5191,\"_174\":5192},\"9d7796132c04\",[],\"“But I quickly observed that [with successive loops], you end up compounding quality issues and you run into these challenges.” The builder’s observations led him to start experimenting with combinations of agentic loops and classical, deterministic AI algorithms, based on a self-taught course of study.\",[],{\"_37\":5195,\"_23\":313,\"_314\":5196,\"_322\":5201,\"_324\":325},\"0291b7718207\",[5197],{\"_37\":5198,\"_23\":318,\"_319\":5199,\"_174\":5200},\"094ef0d4310f\",[],\"“I've used my computer science degree more in the last two years than I have in my entire career, which is just so much fun. So I’m not ‘anti-AI’ at all. What I want to put forward is something that lets you use both modes of tooling in the same context, instead of having a ‘developer mentality’ that requires a hard switch between ‘Here’s an agent that has an LLM as its brain’ versus an agent that has a finite state machine as its brain, and you have to build a bridge between the two. Jido makes both into first-class citizens. I knew as an engineer, as an architect, I would want both.”\",[],{\"_37\":5203,\"_23\":313,\"_314\":5204,\"_322\":5209,\"_324\":408},\"8b84eea66041\",[5205],{\"_37\":5206,\"_23\":318,\"_319\":5207,\"_174\":5208},\"66b38a3b5c20\",[],\"How Can Agents Master Long-Horizon Jobs?\",[],{\"_37\":5211,\"_23\":313,\"_314\":5212,\"_322\":5226,\"_324\":325},\"47dc737f6c1e\",[5213,5217,5222],{\"_37\":5214,\"_23\":318,\"_319\":5215,\"_174\":5216},\"867bc7833538\",[],\"Hostetler admits to reading, and speculating about, the news headlines about agents seemingly getting closer to the holy grail of succeeding on complicated, long-horizon tasks in a single, zero-shot run. “Yes, I read the headlines, and I see the \",{\"_37\":5218,\"_23\":318,\"_319\":5219,\"_174\":5221},\"06d1b38bf227\",[5220],\"2ea1dd0e9cf9\",\"METR reports\",{\"_37\":5223,\"_23\":318,\"_319\":5224,\"_174\":5225},\"8b4acaa0ac87\",[],\". I would say I have hypotheses about the topic, but I acknowledge the limits of my hypotheses.”\",[5227],{\"_37\":5220,\"_23\":49,\"_439\":5228},\"https://metr.org/time-horizons/\",{\"_37\":5230,\"_23\":313,\"_314\":5231,\"_322\":5245,\"_324\":325},\"8d0acac52e79\",[5232,5236,5241],{\"_37\":5233,\"_23\":318,\"_319\":5234,\"_174\":5235},\"2c71dc2a8678\",[],\"“We see new model drops like the [recalled-and-re-released] \",{\"_37\":5237,\"_23\":318,\"_319\":5238,\"_174\":5240},\"45c1bbef0be9\",[5239],\"f0386e204d9f\",\"Fable 5\",{\"_37\":5242,\"_23\":318,\"_319\":5243,\"_174\":5244},\"1fbdf1ad0724\",[],\", which can reportedly do a long-running task for eight hours, uninterrupted.” The creator notes that it can be hard to separate reality from hype, but notes that foundation labs may have different incentives than the rest of us.\",[5246],{\"_37\":5239,\"_23\":49,\"_439\":5247},\"https://www.anthropic.com/news/fable-mythos-access\",{\"_37\":5249,\"_23\":313,\"_314\":5250,\"_322\":5263,\"_324\":325},\"a2cf64eff7a9\",[5251,5255,5259],{\"_37\":5252,\"_23\":318,\"_319\":5253,\"_174\":5254},\"eaf492689fdb\",[],\"“I would think there is a lot of pressure to embed the capability \",{\"_37\":5256,\"_23\":318,\"_319\":5257,\"_174\":5258},\"92183232c806\",[508],\"behind the walls\",{\"_37\":5260,\"_23\":318,\"_319\":5261,\"_174\":5262},\"8058aab96629\",[],\" of the model; to make Fable better via academic RL, or to have a lot of infrastructure running between the API call and the model itself. You see them pulling in things like their web API tools, and doing everything to make Anthropic a walled garden for any corporate customer.”\",[],{\"_37\":5265,\"_23\":313,\"_314\":5266,\"_322\":5271,\"_324\":325},\"41e706c870c8\",[5267],{\"_37\":5268,\"_23\":318,\"_319\":5269,\"_174\":5270},\"14797e507230\",[],\"Hostetler contrasts the lofty ambitions of foundation labs against his day-to-day work, which requires him to service ERP systems maintained by tech leaders across the country who need reliable performance, day in and day out.\",[],{\"_37\":5273,\"_23\":313,\"_314\":5274,\"_322\":5279,\"_324\":325},\"c04dd4c136d2\",[5275],{\"_37\":5276,\"_23\":318,\"_319\":5277,\"_174\":5278},\"840ee6f95b58\",[],\"The creator suggests starting from a strong harness, but also looking into reliability. “A really concrete step is to start with a very high-quality model harness if I'm going to build my own agent. Yes, admittedly, Elixir’s philosophy is ‘let it crash,’ but you don't really want that to happen in practice. Your focus should be: How do you recover from the crash? How do you deliver high-quality software from that?”\",[],{\"_37\":5281,\"_23\":313,\"_314\":5282,\"_322\":5287,\"_324\":1647},\"5824b8efe53d\",[5283],{\"_37\":5284,\"_23\":318,\"_319\":5285,\"_174\":5286},\"16d38849b2d4\",[],\"“Elixir’s philosophy is ‘let it crash,’ but you don't really want that to happen in practice. Your focus should be: How do you recover from the crash? How do you deliver high-quality software from that?” -Mike Hostetler, Creator/Jido\",[],{\"_37\":5289,\"_23\":313,\"_314\":5290,\"_322\":5295,\"_324\":408},\"5959d9db324a\",[5291],{\"_37\":5292,\"_23\":318,\"_319\":5293,\"_174\":5294},\"9adc46734820\",[],\"Squaring the Circle: Autonomous Agents vs. Reliable Uptime\",[],{\"_37\":5297,\"_23\":313,\"_314\":5298,\"_322\":5303,\"_324\":325},\"b4fc4a69d6f3\",[5299],{\"_37\":5300,\"_23\":318,\"_319\":5301,\"_174\":5302},\"e958458240eb\",[],\"The builder recognizes the gap between LLM-powered agents that often fall down for a variety of reasons and Elixir-based systems like Discord and Pinterest that are expected to provide continuous uptime for millions of users. “Without getting too technical, this really was a key design decision of Jido itself.”\",[],{\"_37\":5305,\"_23\":313,\"_314\":5306,\"_322\":5311,\"_324\":325},\"40ebc9c9d5ad\",[5307],{\"_37\":5308,\"_23\":318,\"_319\":5309,\"_174\":5310},\"9b48defee678\",[],\"Hostetler explains that the project was designed to accommodate both builders of independent, solo agents as well as those building for coordinated, multi-agent use cases. “And also, there’s all the infrastructure concerns for when you go to deploy: Where does it live? How is it persisted? It's persisted with what I consider to be two schools of thought.”\",[],{\"_37\":5313,\"_23\":313,\"_314\":5314,\"_322\":5328,\"_324\":325},\"01a202626a6c\",[5315,5319,5324],{\"_37\":5316,\"_23\":318,\"_319\":5317,\"_174\":5318},\"6a69eed46b2a\",[],\"“The first [persistence story] is that I persist my \",{\"_37\":5320,\"_23\":318,\"_319\":5321,\"_174\":5323},\"be252cd6a312\",[5322],\"e73f4b0caaf3\",\"BEAM system\",{\"_37\":5325,\"_23\":318,\"_319\":5326,\"_174\":5327},\"888079506d54\",[],\" back to a durable store, like a Postgres database which is centralized. No issue there. And I would say that more than half our users use this. The other key story is what I call the fully distributed, fault-tolerant approach for scale that you get into with the really big deployments, such as WhatsApp or Facebook Messenger. Multi-region, multi-clustered...the big stuff.”\",[5329],{\"_37\":5322,\"_23\":49,\"_439\":5330},\"https://en.wikipedia.org/wiki/BEAM_(Erlang_virtual_machine)\",{\"_37\":5332,\"_23\":313,\"_314\":5333,\"_322\":5363,\"_324\":325},\"52380ad616db\",[5334,5338,5342,5346,5350,5354,5359],{\"_37\":5335,\"_23\":318,\"_319\":5336,\"_174\":5337},\"04e742e52dfa\",[],\"“Jido was geared for [large-scale deployments]. We could get into the specifics of how Jido was created to live within the BEAM itself, but it \",{\"_37\":5339,\"_23\":318,\"_319\":5340,\"_174\":5341},\"7f2912ffef49\",[508],\"is\",{\"_37\":5343,\"_23\":318,\"_319\":5344,\"_174\":5345},\"f38721e57e56\",[],\" possible. And there are nuances here, too. People come to the project and they sometimes feel like it's overbuilt. Case in point, when you send a message into a Jido agent, we use a dedicated message envelope called a \",{\"_37\":5347,\"_23\":318,\"_319\":5348,\"_174\":5349},\"360df35b7e3c\",[508],\"Jido signal\",{\"_37\":5351,\"_23\":318,\"_319\":5352,\"_174\":5353},\"51665ddc50c6\",[],\", which, when you're in a single-node use case is really and truly overkill. And yet it's the standard built on the \",{\"_37\":5355,\"_23\":318,\"_319\":5356,\"_174\":5358},\"152bbc74099b\",[5357],\"0abd3bbf8418\",\"CloudEvents spec\",{\"_37\":5360,\"_23\":318,\"_319\":5361,\"_174\":5362},\"ca903e998961\",[],\" (which is independent of, and exists outside of Elixir).”\",[5364],{\"_37\":5357,\"_23\":49,\"_439\":5365},\"https://cloudevents.io/\",{\"_37\":5367,\"_23\":313,\"_314\":5368,\"_322\":5373,\"_324\":325},\"46faea6fef5d\",[5369],{\"_37\":5370,\"_23\":318,\"_319\":5371,\"_174\":5372},\"e670c10a6d1b\",[],\"“We do this because you can't have it both ways: You can't have your typical simple JSON payload alongside a robust, enterprise-grade, transport-agnostic CloudEvents envelope and do the same thing. You've got to pick a lane. And so, Jido picked the ‘robust and scalable’ lane. I do work with some teams that are now asking me about the multi-cluster use case. I’m working toward a demo of a multi-region cluster independent of a durable store with resilient agents.”\",[],{\"_37\":5375,\"_23\":313,\"_314\":5376,\"_322\":5397,\"_324\":325},\"8a6207353488\",[5377,5381,5385,5389,5393],{\"_37\":5378,\"_23\":318,\"_319\":5379,\"_174\":5380},\"775967e75006\",[],\"“To clarify, everything I just discussed was \",{\"_37\":5382,\"_23\":318,\"_319\":5383,\"_174\":5384},\"5cfaaca93e88\",[508],\"outside\",{\"_37\":5386,\"_23\":318,\"_319\":5387,\"_174\":5388},\"97a4f29363fe\",[],\" the agentic loop. \",{\"_37\":5390,\"_23\":318,\"_319\":5391,\"_174\":5392},\"accaff227757\",[508],\"Within\",{\"_37\":5394,\"_23\":318,\"_319\":5395,\"_174\":5396},\"83672844a0e3\",[],\" the agentic loop of determinism versus nondeterminism, there are guardrails within which Jido has to operate. It can't just go do everything because it was meant for the Raspberry Pi use case. Because Elixir and BEAM can natively cluster, you could easily have a deploy to support a use case for something like, let’s say, Internet of Things (IoT).”\",[],{\"_37\":5399,\"_23\":313,\"_314\":5400,\"_322\":5414,\"_324\":325},\"36dc5918428c\",[5401,5405,5410],{\"_37\":5402,\"_23\":318,\"_319\":5403,\"_174\":5404},\"f9f53a4cd67b\",[],\"“For example, you’re going to have a central cluster that then works with Jido agents that are deployed in an IoT setting that talk back to the central cluster, and you'd have two agents communicating with one another. You could have that modality in a larger deployment. As an example, one of the big successes in the Elixir world is the robot control project \",{\"_37\":5406,\"_23\":318,\"_319\":5407,\"_174\":5409},\"30ba83fa35f4\",[5408],\"74bfc2d507f3\",\"Beam Bots\",{\"_37\":5411,\"_23\":318,\"_319\":5412,\"_174\":5413},\"ac05df5ad6aa\",[],\", which adopted Jido natively, so we’re seeing use cases for non-clustered low-level IoT all the way up to your typical Claude Code agent harness.”\",[5415],{\"_37\":5408,\"_23\":49,\"_439\":5416},\"https://github.com/beam-bots\",{\"_37\":5418,\"_23\":313,\"_314\":5419,\"_322\":5424,\"_324\":408},\"6471f5331431\",[5420],{\"_37\":5421,\"_23\":318,\"_319\":5422,\"_174\":5423},\"88eb6436d108\",[],\"How Startup Founders Should Think About Agentic Products\",[],{\"_37\":5426,\"_23\":313,\"_314\":5427,\"_322\":5432,\"_324\":325},\"b02dc08a377e\",[5428],{\"_37\":5429,\"_23\":318,\"_319\":5430,\"_174\":5431},\"71c2612c69e4\",[],\"When asked how he would advise startup founders thinking of launching agentic products, Hostetler offers a handful of suggestions. “For your first level of planning, I would say: Focus on the default. How you define what an ‘agent’ is could be colored by your Claude Code glasses! Make sure that as you're defining an agent in terms of what you bring into your infrastructure, and that you have a tighter, concrete definition of what that is.”\",[],{\"_37\":5434,\"_23\":313,\"_314\":5435,\"_322\":5448,\"_324\":325},\"322cd96a4729\",[5436,5440,5444],{\"_37\":5437,\"_23\":318,\"_319\":5438,\"_174\":5439},\"a7c7ebbd0554\",[],\"“Second, the other mistake that I've seen (and have made personally) is that everybody wants to let the LLM ‘take the wheel.’ At the point when you're designing infrastructure to build a company on, you have to draw clear boundaries between nondeterminism and determinism, and create gates to bring it back and forth. And this is \",{\"_37\":5441,\"_23\":318,\"_319\":5442,\"_174\":5443},\"1ea4bce42d9f\",[508],\"outside of\",{\"_37\":5445,\"_23\":318,\"_319\":5446,\"_174\":5447},\"9510224427a1\",[],\" the agentic loop.”\",[],{\"_37\":5450,\"_23\":313,\"_314\":5451,\"_322\":5456,\"_324\":325},\"e80672b9d929\",[5452],{\"_37\":5453,\"_23\":318,\"_319\":5454,\"_174\":5455},\"61e9e3aed2a8\",[],\"The creator suggests that customer expectations for agentic projects have grown over time. “After a few years, we’ve seen the patterns that ‘work,’ whether it's natural language or dynamic UI, we know the patterns to create user value and what users expect. The innovation has shifted from, ‘We can do just about anything,’ to ‘The user expectations are pretty much there, so how do you deliver on that?’”\",[],{\"_37\":5458,\"_23\":313,\"_314\":5459,\"_322\":5464,\"_324\":325},\"32013237d35f\",[5460],{\"_37\":5461,\"_23\":318,\"_319\":5462,\"_174\":5463},\"95b20553d1ab\",[],\"Hostetler describes an interesting project a friend built in Elixir that effectively replicated Claude Code’s functionality, but for posting on LinkedIn, using a dedicated agent with a specific ID. “Persistence wasn't dynamically allocated, but the agent would persist, and then it would hibernate and rehydrate when you wanted to edit a piece of content.”\",[],{\"_37\":5466,\"_23\":313,\"_314\":5467,\"_322\":5479,\"_324\":325},\"3309765c1194\",[5468,5472,5476],{\"_37\":5469,\"_23\":318,\"_319\":5470,\"_174\":5471},\"6b38e3208d9b\",[],\"The project let users edit posts normally via the LinkedIn GUI, but also supported giving instructions to a coordinator agent. “The key is, he made that persistent throughout the entire experience. So, no matter which modality you were interacting with, there was a kind of predefined team structure for how the agents would interact with your content. And it just struck me as this really innovative mental leap, that any agentic products I see \",{\"_37\":5473,\"_23\":318,\"_319\":5474,\"_174\":5475},\"302d71f6b276\",[508],\"aren't thinking big enough\",{\"_37\":5477,\"_23\":318,\"_319\":5478,\"_174\":1518},\"5e2c8479951c\",[],[],{\"_37\":5481,\"_23\":313,\"_314\":5482,\"_322\":5487,\"_324\":325},\"f9024ac9c07d\",[5483],{\"_37\":5484,\"_23\":318,\"_319\":5485,\"_174\":5486},\"090d9b48e200\",[],\"The creator suggests that the main challenge he’d offer to startup founders and product owners is to think bigger. “I don't see a lot of product owners thinking about [larger ideas] because they're really stuck in the mindset of, ‘Claude Code is my agent. It can do everything. I have to always use Claude Code.’ I say: ‘No, you really need to break out of that mental model.’”\",[],{\"_37\":5489,\"_23\":313,\"_314\":5490,\"_322\":5495,\"_324\":1647},\"d279075f7c24\",[5491],{\"_37\":5492,\"_23\":318,\"_319\":5493,\"_174\":5494},\"1407b21e0ca6\",[],\"“It just struck me...that any agentic products I see aren't thinking big enough. I don't see a lot of product owners thinking about [larger ideas] because they're really stuck in the mindset of, ‘Claude Code is my agent. I say: ‘No, you really need to break out of that mental model.’”\",[],{\"_37\":5497,\"_23\":313,\"_314\":5498,\"_322\":5503,\"_324\":408},\"8bd22d228d02\",[5499],{\"_37\":5500,\"_23\":318,\"_319\":5501,\"_174\":5502},\"febccbe3dad6\",[],\"How to Think About Enterprise Adoption\",[],{\"_37\":5505,\"_23\":313,\"_314\":5506,\"_322\":5520,\"_324\":325},\"55d44d50b054\",[5507,5511,5516],{\"_37\":5508,\"_23\":318,\"_319\":5509,\"_174\":5510},\"140634326dd3\",[],\"When asked how he would advise ambitious startup founders looking to sell AI products to enterprises, Hostetler draws upon his own day-to-day experience. “I would say that we're early. I would acknowledge the timeline displacement, \",{\"_37\":5512,\"_23\":318,\"_319\":5513,\"_174\":5515},\"60c8c804bc83\",[5514],\"395e4914b63e\",\"Ray Kurzweil’s\",{\"_37\":5517,\"_23\":318,\"_319\":5518,\"_174\":5519},\"940cd7aa0849\",[],\" riff on the William Gibson quote about the future being here, just not evenly distributed.”\",[5521],{\"_37\":5514,\"_23\":49,\"_439\":5522},\"https://www.writingsbyraykurzweil.com/the-law-of-accelerating-returns\",{\"_37\":5524,\"_23\":313,\"_314\":5525,\"_322\":5546,\"_324\":325},\"69e3ce55443f\",[5526,5530,5534,5538,5542],{\"_37\":5527,\"_23\":318,\"_319\":5528,\"_174\":5529},\"c2cbce688847\",[],\"“What's happening in enterprises is that there are engineering teams who \",{\"_37\":5531,\"_23\":318,\"_319\":5532,\"_174\":5533},\"8f43979ee05a\",[508],\"are \",{\"_37\":5535,\"_23\":318,\"_319\":5536,\"_174\":5537},\"7386ba0066ca\",[],\"using the latest tools, and they're on that curve somewhere. But \",{\"_37\":5539,\"_23\":318,\"_319\":5540,\"_174\":5541},\"80dac42e9497\",[508],\"the operations teams\",{\"_37\":5543,\"_23\":318,\"_319\":5544,\"_174\":5545},\"f3f8661263be\",[],\" are just starting.” The creator notes that he works with operations teams that are just starting to encounter the same challenges that engineering teams faced 18 months ago: Worrying whether AI will replace them, realizing AI likely won’t replace them, then trying to figure out how to fit into this brave new world.\",[],{\"_37\":5548,\"_23\":313,\"_314\":5549,\"_322\":5570,\"_324\":325},\"d6bdea3ba045\",[5550,5554,5559,5563,5567],{\"_37\":5551,\"_23\":318,\"_319\":5552,\"_174\":5553},\"6dbb3fee6f18\",[],\"“We're early in that phase [where people are trying to find] pragmatic use cases. The issue isn't so much the tools themselves. What's interesting to me is the layer above them, which I wrote about in \",{\"_37\":5555,\"_23\":318,\"_319\":5556,\"_174\":5558},\"e44831398a0a\",[5557],\"ba5da2cbf968\",\"my blog on orchestration\",{\"_37\":5560,\"_23\":318,\"_319\":5561,\"_174\":5562},\"3c012b7e05ac\",[],\". To me, orchestration is the cornerstone, because there's nothing actually orchestrating \",{\"_37\":5564,\"_23\":318,\"_319\":5565,\"_174\":5566},\"c493d94c2769\",[508],\"the tools\",{\"_37\":5568,\"_23\":318,\"_319\":5569,\"_174\":1518},\"0ddedf4c18ed\",[],[5571],{\"_37\":5557,\"_23\":49,\"_439\":5572},\"https://mike-hostetler.com/blog/orchestrating-intelligence-why-every-company-will-have-a-chief-agent-officer-y4EQpjr9\",{\"_37\":5574,\"_23\":313,\"_314\":5575,\"_322\":5580,\"_324\":325},\"ffdaf5b669b5\",[5576],{\"_37\":5577,\"_23\":318,\"_319\":5578,\"_174\":5579},\"d9fbf1dd4f20\",[],\"“I think that corporations are set up in silos. Sales isn't talking to marketing, who isn't talking to ops. Ops isn't talking to product. So, the pragmatic things that we need are places to deploy AI experiments, because the thing we've emphasized internally isn't so much AI as an end state, but rather, AI as iteration.”\",[],{\"_37\":5582,\"_23\":313,\"_314\":5583,\"_322\":5605,\"_324\":325},\"86f97ec55b10\",[5584,5588,5593,5597,5601],{\"_37\":5585,\"_23\":318,\"_319\":5586,\"_174\":5587},\"9aad627e9643\",[],\"The creator recounts a story of a team he works with using Claude Code to build a dashboard. In a completely unintended consequence, Claude Code used the Google Workspace connect to deploy the dashboard to Google Sheets via \",{\"_37\":5589,\"_23\":318,\"_319\":5590,\"_174\":5592},\"865280795f71\",[5591],\"2ed51292aa7f\",\"Apps Script\",{\"_37\":5594,\"_23\":318,\"_319\":5595,\"_174\":5596},\"4e0143f6f2a2\",[],\" to create a public URL via, of all things, a Google Sheets backdoor. “I burst out laughing. There's an entire product \",{\"_37\":5598,\"_23\":318,\"_319\":5599,\"_174\":5600},\"25776a76f3df\",[508],\"right there,\",{\"_37\":5602,\"_23\":318,\"_319\":5603,\"_174\":5604},\"8a472903f3c6\",[],\" in that space between a non-technical user and a model’s instincts to propose unexpected solutions.”\",[5606],{\"_37\":5591,\"_23\":49,\"_439\":5607},\"https://developers.google.com/apps-script/overview\",{\"_37\":5609,\"_23\":313,\"_314\":5610,\"_322\":5615,\"_324\":325},\"1afc27d42096\",[5611],{\"_37\":5612,\"_23\":318,\"_319\":5613,\"_174\":5614},\"19bdb5bfad0f\",[],\"Hostetler also points out a strong business opportunity in governance. “During my tour of duty in corporate, I spent two years as a VP of enterprise technology. I managed all the IT and compliance stuff for a publicly-traded company.” The creator notes that “AI governance” could easily go beyond basic compliance frameworks like SOC 2.\",[],{\"_37\":5617,\"_23\":313,\"_314\":5618,\"_322\":5623,\"_324\":325},\"8fe6ad610c6c\",[5619],{\"_37\":5620,\"_23\":318,\"_319\":5621,\"_174\":5622},\"9e450b53590e\",[],\"“Every LLM call in your organization will eventually be proxied and logged. But nobody's doing this at the moment. Think about just capturing it as one problem. Analyzing and fingerprinting.” The creator predicts that sooner than we think, organizations may need to be “AI compliant” with whatever new version of the SOC2 standard that arises.\",[],{\"_37\":5625,\"_23\":313,\"_314\":5626,\"_322\":5631,\"_324\":325},\"ef3eab6b09a3\",[5627],{\"_37\":5628,\"_23\":318,\"_319\":5629,\"_174\":5630},\"856e44ea47e3\",[],\"“Another [enterprise use case] I’d look at is insurance. Cybersecurity insurance with AI, mark my words, in the next two years, that will be a ‘thing.’ And this won’t just be your CrowdStrike-style, on-device stuff. It's all the things that get exfiltrated from what you build. Nobody's talking about that. The strategy there is very immature. Admittedly, this isn’t a very sexy space, but I think there's a lot of money to be made there.”\",[],{\"_37\":5633,\"_23\":313,\"_314\":5634,\"_322\":5639,\"_324\":408},\"8c5c46a705f3\",[5635],{\"_37\":5636,\"_23\":318,\"_319\":5637,\"_174\":5638},\"e55bcf804130\",[],\"The Future of Elixir and Durable Systems\",[],{\"_37\":5641,\"_23\":313,\"_314\":5642,\"_322\":5647,\"_324\":325},\"44f07555baa8\",[5643],{\"_37\":5644,\"_23\":318,\"_319\":5645,\"_174\":5646},\"83aa3f705283\",[],\"The creator is realistic about how his project and the greater Elixir ecosystem figure into the future of software development. “A non-Elixir shop is going to have a hard time adopting Elixir. So I can see Jido becoming an agentic sidecar that eventually makes it easy for TypeScript devs to spin up an agent with Elixir-level durability.”\",[],{\"_37\":5649,\"_23\":313,\"_314\":5650,\"_322\":5664,\"_324\":325},\"dbad934515e7\",[5651,5655,5660],{\"_37\":5652,\"_23\":318,\"_319\":5653,\"_174\":5654},\"ddf9d15e994c\",[],\"“This might be for a market segment that could be ‘one step below’ the \",{\"_37\":5656,\"_23\":318,\"_319\":5657,\"_174\":5659},\"7929d2ffed1a\",[5658],\"f82cbe3cdbf8\",\"Temporals\",{\"_37\":5661,\"_23\":318,\"_319\":5662,\"_174\":5663},\"82c9fc0d6074\",[],\" of the world. To clarify, I’ve used Temporal before, and it’s fantastic, though expensive. But thinking pragmatically, there are a lot more TypeScript devs in the world. Python will be its own thing. TypeScript devs are going to need an SDK to easily call Jido agents, and then Jido agents will operate async, and it'll just be self-contained with a dashboard to spin it up.”\",[5665],{\"_37\":5658,\"_23\":49,\"_439\":5666},\"https://temporal.io/\",{\"_37\":5668,\"_23\":313,\"_314\":5669,\"_322\":5683,\"_324\":325},\"b3f069861ce0\",[5670,5674,5679],{\"_37\":5671,\"_23\":318,\"_319\":5672,\"_174\":5673},\"7dc862f60a3b\",[],\"The creator acknowledges that popularly-used languages and frameworks like TypeScript and \",{\"_37\":5675,\"_23\":318,\"_319\":5676,\"_174\":5678},\"a5ab9089e32a\",[5677],\"cc95782eb602\",\"Node.js\",{\"_37\":5680,\"_23\":318,\"_319\":5681,\"_174\":5682},\"05d94911aeac\",[],\" are excellent for what they are, and that they solve specific problems. “But you need to be able to spin things up inexpensively that companies built on TypeScript can solve either by running in Docker via their infrastructure, or run out of a cloud where they can spin up platforms of agents, with one agent per customer (or dozens of agents per customer).”\",[5684],{\"_37\":5677,\"_23\":49,\"_439\":5685},\"http://node.js\",{\"_37\":5687,\"_23\":313,\"_314\":5688,\"_322\":5693,\"_324\":325},\"1c459bacc235\",[5689],{\"_37\":5690,\"_23\":318,\"_319\":5691,\"_174\":5692},\"4773952f0d8b\",[],\"“And to be able to do that practically is just going to be overwhelmingly expensive, even with Cloud costs. And you need to be able to stitch it all together, potentially using non-deterministic agents with deterministic agents to deliver a strong product experience. It’s a very early picture, but that's where I see the market now.”\",[],19,\"https://cdn.sanity.io/images/50q6fr1p/production/a119969ab98a18b0937559865cbc6ee4dee53c2d-1200x630.png\",\"39be2ceb-4600-40c4-80a9-dc0c7b9504dd\",\"2026-08-04T16:24:00.000Z\",\"how-to-make-agents-durable-for-concurrent-systems\",\"How to Make Agents Durable for Concurrent Systems\",[5701,5703,5705],{\"_37\":5702,\"_116\":3618,\"_23\":118},\"8MHMqhfAkJQ7\",{\"_37\":5704,\"_116\":3677,\"_23\":118},\"v08UzOpf6wKa\",{\"_37\":5706,\"_116\":4112,\"_23\":118},\"hOpKZoYNyuau\",\"outlines\",[5709,5714,5719,5724,5729,5734,5739,5744,5749,5754,5759],{\"_37\":2056,\"_23\":313,\"_314\":5710,\"_322\":5713,\"_324\":408},[5711],{\"_37\":2059,\"_23\":318,\"_319\":5712,\"_174\":2061},[],[],{\"_37\":2208,\"_23\":313,\"_314\":5715,\"_322\":5718,\"_324\":408},[5716],{\"_37\":2211,\"_23\":318,\"_319\":5717,\"_174\":2213},[],[],{\"_37\":2313,\"_23\":313,\"_314\":5720,\"_322\":5723,\"_324\":449},[5721],{\"_37\":2316,\"_23\":318,\"_319\":5722,\"_174\":2318},[],[],{\"_37\":2361,\"_23\":313,\"_314\":5725,\"_322\":5728,\"_324\":408},[5726],{\"_37\":2364,\"_23\":318,\"_319\":5727,\"_174\":2366},[],[],{\"_37\":2424,\"_23\":313,\"_314\":5730,\"_322\":5733,\"_324\":449},[5731],{\"_37\":2427,\"_23\":318,\"_319\":5732,\"_174\":2429},[],[],{\"_37\":2472,\"_23\":313,\"_314\":5735,\"_322\":5738,\"_324\":408},[5736],{\"_37\":2475,\"_23\":318,\"_319\":5737,\"_174\":2477},[],[],{\"_37\":2523,\"_23\":313,\"_314\":5740,\"_322\":5743,\"_324\":449},[5741],{\"_37\":2526,\"_23\":318,\"_319\":5742,\"_174\":2528},[],[],{\"_37\":2587,\"_23\":313,\"_314\":5745,\"_322\":5748,\"_324\":408},[5746],{\"_37\":2590,\"_23\":318,\"_319\":5747,\"_174\":2592},[],[],{\"_37\":2672,\"_23\":313,\"_314\":5750,\"_322\":5753,\"_324\":449},[5751],{\"_37\":2675,\"_23\":318,\"_319\":5752,\"_174\":2677},[],[],{\"_37\":2755,\"_23\":313,\"_314\":5755,\"_322\":5758,\"_324\":408},[5756],{\"_37\":2758,\"_23\":318,\"_319\":5757,\"_174\":2760},[],[],{\"_37\":2781,\"_23\":313,\"_314\":5760,\"_322\":5763,\"_324\":408},[5761],{\"_37\":2784,\"_23\":318,\"_319\":5762,\"_174\":2786},[],[],\"2023-10-12T22:08:00.000Z\",{\"_147\":5766,\"_23\":295,\"_28\":2829,\"_298\":196,\"_299\":196,\"_32\":5769},{\"_23\":149,\"_114\":5767},{\"_116\":5768,\"_23\":118},\"image-b67963a5a553d32ba4c307eb7425f753bcdaba6a-800x300-jpg\",\"How Does Generative AI Work with Devops and Incident Response?\",{\"_23\":366,\"_368\":5771},\"generative-ai-incident-response-devops\",[5773,5778,5786,5794,5799,5806],{\"_11\":5774,\"_13\":927,\"_14\":5775,\"_23\":5776,\"_24\":5774,\"_292\":5777},\"2022-11-10T14:12:09Z\",\"NgvOL1OJzLJ5agOL2f1xw3\",\"tag\",\"Artificial Intelligence (AI)\",{\"_11\":5779,\"_13\":5780,\"_14\":5781,\"_23\":5776,\"_24\":5779,\"_28\":5782,\"_292\":5783,\"_366\":5784},\"2023-07-17T23:43:05Z\",\"4bf77923-361f-49e2-a0ed-07d84a6b3dc4\",\"TYN2nJHhdg1RiOg5ta15Uk\",\"Coding assistants are AI-powered tools that help developers create code.\",\"Coding Assistants\",{\"_23\":366,\"_368\":5785},\"coding assistants\",{\"_11\":5787,\"_13\":5788,\"_14\":5789,\"_23\":5776,\"_24\":5787,\"_28\":5790,\"_292\":5791,\"_366\":5792},\"2023-07-17T23:27:35Z\",\"5893abdf-ea1d-4456-8114-7493984631a6\",\"NgvOL1OJzLJ5agOL2f1ypV\",\"Large-language models are algorithms that, when trained on datasets, power AI chatbots to return results when prompted.\",\"LLM\",{\"_23\":366,\"_368\":5793},\"llm\",{\"_11\":5795,\"_13\":5796,\"_14\":5797,\"_23\":5776,\"_24\":5795,\"_292\":5798},\"2022-08-23T19:28:33Z\",\"7e93d635-2341-4db2-9cb6-141475b9232c\",\"410YivtbVgNW6RgQxPP59n\",\"Incident Response\",{\"_11\":5800,\"_13\":936,\"_14\":5801,\"_23\":5776,\"_24\":5800,\"_28\":5802,\"_292\":5803,\"_366\":5804},\"2023-08-28T22:38:33Z\",\"NgvOL1OJzLJ5agOL2f1pgf\",\"Generative AI is a form of machine learning that produces content (such as text, images, video, or audio) in response to being prompted.\",\"Generative AI\",{\"_23\":366,\"_368\":5805},\"generative-ai\",{\"_11\":5807,\"_13\":5808,\"_14\":5809,\"_23\":5776,\"_24\":5807,\"_28\":3541,\"_292\":5810,\"_366\":5811},\"2022-09-09T00:13:21Z\",\"v6woxsuuKgC7emJ9vwJKve\",\"aeN0todGvIjkWyTywFzd0a\",\"DevOps\",{\"_23\":366,\"_368\":5812},\"devops\",\"Incident Response and DevOps in the Age of Generative AI\",\"preview\",\"actionData\",\"errors\"]\n");</script><!--$?--><template id="B:1"></template><!--/$--></div><script>function $RC(a,b){a=document.getElementById(a);b=document.getElementById(b);b.parentNode.removeChild(b);if(a){a=a.previousSibling;var f=a.parentNode,c=a.nextSibling,e=0;do{if(c&&8===c.nodeType){var d=c.data;if("/$"===d)if(0===e)break;else e--;else"$"!==d&&"$?"!==d&&"$!"!==d||e++}d=c.nextSibling;f.removeChild(c);c=d}while(c);for(;b.firstChild;)f.insertBefore(b.firstChild,c);a.data="$";a._reactRetry&&a._reactRetry()}};$RC("B:0","S:0")</script><div hidden id="S:1"><script>window.__reactRouterContext.streamController.close();</script></div><script>$RC("B:1","S:1")</script><!-- This script is automatically inserted by Netlify for Real User Monitoring (RUM). -->
<script async id="netlify-rum-container" src="/.netlify/scripts/rum" data-netlify-cwv-token="eyJhbGciOiJIUzI1NiIsInR5cCI6IkpXVCJ9.eyJzaXRlX2lkIjoiYTdjMTM3MTktMWUxMC00ZmI2LWJjOTAtMmZkMDczYTM1NzIwIiwiYWNjb3VudF9pZCI6IjU5MDI2NDM5Y2YzMjFjNzM1OWU5OGMxZiIsImRlcGxveV9pZCI6IjZhYTQ5MWRhYjdhNzU0MDAwOGI3ZGNlZCIsImlzcyI6Im5ldGxpZnkifQ.X3CF6DFeTNXTjNx0HofPwoME_M2yzpJLUB5QdF-R0Xg"></script>