--- title: System one models and Jev: what engineers showed at the... description: System one models were the central theme at the AI Engineer Code Summit. Unlike generative models, these return a structured choice or score instead of... canonical: https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: System one models and Jev: what engineers showed at the AI Engineer Code Summit | daily.dev og:description: System one models were the central theme at the AI Engineer Code Summit. Unlike generative models, these return a structured choice or score instead of... og:url: https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg og:image: https://api.daily.dev/og/posts/WxoUqfDDG.png og:image:alt: System one models and Jev: what engineers showed at the AI Engineer Code Summit og:image:width: 1200 og:image:height: 630 og:locale: en --- > ## Documentation Index > Fetch the complete documentation index at: https://daily.dev/llms.txt > Use this file to discover all available pages before exploring further. # System one models and Jev: what engineers showed at the AI Engineer Code Summit **[Collections](https://daily.dev/sources/collections)** · 3 min read · 0 upvotes · 0 comments ## Summary System one models were the central theme at the AI Engineer Code Summit. Unlike generative models, these return a structured choice or score instead of free-form text. Jev, released in September and led by Diogo Almeida of TypeSafe AI, kicked off the trend, pitched as 'reliable intelligence' for calls embedded directly in normal code. Glean uses Jev for zero-shot decision tasks without retraining. UiPath ran a document-ranking tournament with thousands of Jev calls finishing in under a second. Workday prototyped a self-filling time-off form. @swyx and the Latent Space podcast called it the start of a 'System One revolution.' ## Content System one models were the main theme at the AI Engineer Code Summit. These are non-generative models. Instead of free-form text, they return a structured choice or a score. Jev, released in September by TypeSafe AI, kicked off the trend. Its lead creator is Diogo Almeida, TypeSafe AI's co-founder and CEO, who calls it "reliable intelligence" for "automating everything" and argues that code should make system one calls directly. @swyx, relaying the Latent Space podcast's coverage, summed up the mood: the System One revolution is upon us. ## How it works Unlike a chat model, Jev takes a state plus a narrow yes/no or multiple-choice question. It returns a fast, structured answer with a calibrated probability. There's no prose to parse. ## What people are building - Glean uses Jev for zero-shot decision tasks, with no retraining. - UiPath ran a document-ranking tournament. Thousands of Jev calls finished in under a second. - Workday prototyped a time-off form that fills itself in as the user types. ## A real pipeline: DoltHub's sales leads DoltHub put Jev into its internal sales-leads pipeline to flag organizations that are internet service providers or hosting companies. It asked yes/no questions about organization names and reverse DNS hostnames. TypeSafe suggests treating scores above 0.7 as yes and below 0.3 as no. The middle band is where it gets awkward. In one example, the ISP and hosting scores landed between 0.3 and 0.59, so the pipeline kept an organization that really was an ISP. DoltHub says Jev still catches many others it used to miss, so I'd call that an acceptable miss rather than a flaw. But you do need a plan for the ambiguous scores. ## An open-weight challenger A test on 661 real agent tool calls compared Jev with d3, an open-weight decision model (26B parameters, Apache-2.0). The task was deciding which calls need human review. | | d3 | Jev | |---|---|---| | Calls sent to human review | 8% | 50% | | AUROC | 0.989 | 0.912 | | Label agreement | 87% | 78% | | Planted prompt injections caught | 87% | 100% | | Review cost per decision | $0.039 | $0.252 | d3 ranked calls better and sent far fewer to a human. Jev was stronger on prompt injection, catching every planted attack. Self-hosting d3 breaks even against review costs at roughly 500 decisions a day. Which one wins depends on what hurts more: reviewer time or a missed attack. If injection is your main threat, Jev's 100% matters. If review volume is the bottleneck, half of all calls going to a human is hard to live with. ## Questions this post answers ### What is a system one model in AI engineering? A system one model is a non-generative model that returns a structured choice or score instead of writing free-form text. These models are designed to be called directly from normal program code, functioning as fast, reliable decision points rather than conversational chatbots. Jev, released in September by TypeSafe AI co-founder Diogo Almeida, is an early example driving this approach. _Developers evaluating structured-output models over chatbots can follow this shift on daily.dev._ ### What is Jev used for in production applications? Jev is used for fast, structured decision-making tasks embedded in code rather than generating conversational text. Glean uses it for zero-shot decision tasks without retraining per task, UiPath ran a document-ranking tournament completing thousands of Jev calls in under a second, and Workday prototyped a time-off form that auto-fills as a user types. _Teams weighing structured AI calls against chat-based models can track real examples like these on daily.dev._ ## Similar posts on daily.dev - [Decision AI: The next level in agentic system architecture](https://daily.dev/posts/decision-ai-the-next-level-in-agentic-system-architecture-rvqpmyltg) · Medium · 2 upvotes · 0 comments - [What is Jev? How System One models could change AI agents : fireup.pro](https://daily.dev/posts/what-is-jev-how-system-one-models-could-change-ai-agents-fireup-pro-eqcqub8on) · FireUp · 1 upvotes · 0 comments --- Tags: [#ai-engineering](https://daily.dev/tags/ai-engineering), [#jev](https://daily.dev/tags/jev) [View this post on daily.dev](https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg) ```json {"@context":"https://schema.org","@graph":[{"@type":"Organization","@id":"https://daily.dev/#organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180},"sameAs":["https://twitter.com/dailydotdev","https://github.com/dailydotdev","https://www.linkedin.com/company/daily-dev-ltd"]},{"@type":"WebSite","@id":"https://daily.dev/#website","url":"https://daily.dev","name":"daily.dev","publisher":{"@id":"https://daily.dev/#organization"},"potentialAction":{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https://daily.dev/search?q={search_term_string}"},"query-input":"required name=search_term_string"}}]} {"@context":"https://schema.org","@type":"TechArticle","headline":"System one models and Jev: what engineers showed at the AI Engineer Code Summit","url":"https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg"},"datePublished":"2026-11-12T05:31:01.621Z","dateModified":"2026-11-13T22:17:19.766Z","description":"System one models were the central theme at the AI Engineer Code Summit. Unlike generative models, these return a structured choice or score instead of...","isAccessibleForFree":true,"articleSection":"Collections","inLanguage":"en","publisher":{"@type":"Organization","name":"daily.dev","url":"https://daily.dev","logo":{"@type":"ImageObject","url":"https://daily.dev/apple-touch-icon.png","width":180,"height":180}},"author":{"@type":"Organization","name":"Collections","logo":"https://media.daily.dev/image/upload/s--fk_6ycEi--/f_auto,q_auto/v1780996001/logos/collections?_a=BAMAMiWQ0","url":"https://daily.dev/sources/collections"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"ai-engineering,jev","timeRequired":"PT3M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"Collections","item":"https://daily.dev/sources/collections"},{"@type":"ListItem","position":3,"name":"System one models and Jev: what engineers showed at the AI Engineer Code Summit"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/system-one-models-and-jev-what-engineers-showed-at-the-ai-engineer-code-summit-wxouqfddg#faq","mainEntity":[{"@type":"Question","name":"What is a system one model in AI engineering?","acceptedAnswer":{"@type":"Answer","text":"A system one model is a non-generative model that returns a structured choice or score instead of writing free-form text. These models are designed to be called directly from normal program code, functioning as fast, reliable decision points rather than conversational chatbots. Jev, released in September by TypeSafe AI co-founder Diogo Almeida, is an early example driving this approach. Developers evaluating structured-output models over chatbots can follow this shift on daily.dev."}},{"@type":"Question","name":"What is Jev used for in production applications?","acceptedAnswer":{"@type":"Answer","text":"Jev is used for fast, structured decision-making tasks embedded in code rather than generating conversational text. Glean uses it for zero-shot decision tasks without retraining per task, UiPath ran a document-ranking tournament completing thousands of Jev calls in under a second, and Workday prototyped a time-off form that auto-fills as a user types. Teams weighing structured AI calls against chat-based models can track real examples like these on daily.dev."}}]} ```