--- title: Why I'm excited about Jev-like classifiers | daily.dev description: An exploration of why zero-shot classifiers like Jev are appealing as a middle step between raw LLM prompting and fully trained custom classifiers. The piece... canonical: https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i twitter:card: summary_large_image twitter:site: @dailydotdev og:type: website og:site_name: daily.dev og:title: Why I'm excited about Jev-like classifiers | daily.dev og:description: An exploration of why zero-shot classifiers like Jev are appealing as a middle step between raw LLM prompting and fully trained custom classifiers. The piece... og:url: https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i og:image: https://api.daily.dev/og/posts/9TcsyLH6i.png og:image:alt: Why I'm excited about Jev-like classifiers 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. # Why I'm excited about Jev-like classifiers **[swizec.com](https://daily.dev/sources/swizec)** · 4 min read · 0 upvotes · 0 comments ## Summary An exploration of why zero-shot classifiers like Jev are appealing as a middle step between raw LLM prompting and fully trained custom classifiers. The piece explains how Jev-style tools post-train an LLM to output a single token/choice instead of generating full text, making them fast enough for real-time use cases like a dinosaur-jumping game demo. A five-step roadmap is proposed for building production AI products: start with an LLM, gather real-world data, formalize a rubric, iterate for accuracy, then train a custom classifier once the problem is well understood. ## Full article daily.dev links to this article rather than hosting it. Read it at the original source: ## Questions this post answers ### What is a zero-shot classifier like Jev and how is it different from calling an LLM directly? A zero-shot classifier is an LLM post-trained to output a single token representing a decision from a fixed set of choices, rather than generating free text through a chain of reasoning. This makes it dramatically faster and cheaper than a general LLM call while still requiring no training data or evals, just prompt engineering, since it can make decisions in roughly 100 milliseconds, fast enough to drive a real-time game. _Developers weighing LLM prompting against faster classifier approaches can track this comparison on daily.dev._ ### What are the steps to go from an LLM prototype to a production-ready classifier for a business use case? Start by building the feature with a general LLM and iterating on the prompt, then begin collecting a real-world dataset from that usage. Once the prompt reliably works, turn it into a clear rubric usable by a zero-shot classifier like Jev, keep iterating for accuracy, and finally use the accumulated real-world evals to train a custom classifier fine-tuned to the specific task for speed, scale, and precision. _Teams planning an AI feature roadmap can follow this kind of staged approach found via daily.dev._ --- Tags: [#machine-learning](https://daily.dev/tags/machine-learning), [#llm](https://daily.dev/tags/llm), [#prompt-engineering](https://daily.dev/tags/prompt-engineering) [View this post on daily.dev](https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i) ```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":"Why I'm excited about Jev-like classifiers","url":"https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i","mainEntityOfPage":{"@type":"WebPage","@id":"https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i"},"datePublished":"2026-09-29T15:48:06.118Z","dateModified":"2026-09-29T19:41:42.712Z","description":"An exploration of why zero-shot classifiers like Jev are appealing as a middle step between raw LLM prompting and fully trained custom classifiers. The piece...","image":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/5306a45bfc90512b8e7e3726c73fbe7d?_a=AQAEuop","thumbnailUrl":"https://media.daily.dev/image/upload/f_auto,q_auto/v1/posts/5306a45bfc90512b8e7e3726c73fbe7d?_a=AQAEuop","isAccessibleForFree":true,"articleSection":"swizec.com","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":"swizec.com","logo":"https://media.daily.dev/image/upload/t_logo,f_auto/v1/logos/a0733daf6e014e23a638ded7f2d4cd70","url":"https://daily.dev/sources/swizec"},"commentCount":0,"discussionUrl":"https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i","interactionStatistic":[{"@type":"InteractionCounter","interactionType":{"@type":"LikeAction"},"userInteractionCount":0},{"@type":"InteractionCounter","interactionType":{"@type":"CommentAction"},"userInteractionCount":0}],"keywords":"machine-learning,llm,prompt-engineering","timeRequired":"PT4M"} {"@context":"https://schema.org","@type":"BreadcrumbList","itemListElement":[{"@type":"ListItem","position":1,"name":"Home","item":"https://daily.dev"},{"@type":"ListItem","position":2,"name":"swizec.com","item":"https://daily.dev/sources/swizec"},{"@type":"ListItem","position":3,"name":"Why I'm excited about Jev-like classifiers"}]} {"@context":"https://schema.org","@type":"FAQPage","@id":"https://daily.dev/posts/why-i-m-excited-about-jev-like-classifiers-9tcsylh6i#faq","mainEntity":[{"@type":"Question","name":"What is a zero-shot classifier like Jev and how is it different from calling an LLM directly?","acceptedAnswer":{"@type":"Answer","text":"A zero-shot classifier is an LLM post-trained to output a single token representing a decision from a fixed set of choices, rather than generating free text through a chain of reasoning. This makes it dramatically faster and cheaper than a general LLM call while still requiring no training data or evals, just prompt engineering, since it can make decisions in roughly 100 milliseconds, fast enough to drive a real-time game. Developers weighing LLM prompting against faster classifier approaches can track this comparison on daily.dev."}},{"@type":"Question","name":"What are the steps to go from an LLM prototype to a production-ready classifier for a business use case?","acceptedAnswer":{"@type":"Answer","text":"Start by building the feature with a general LLM and iterating on the prompt, then begin collecting a real-world dataset from that usage. Once the prompt reliably works, turn it into a clear rubric usable by a zero-shot classifier like Jev, keep iterating for accuracy, and finally use the accumulated real-world evals to train a custom classifier fine-tuned to the specific task for speed, scale, and precision. Teams planning an AI feature roadmap can follow this kind of staged approach found via daily.dev."}}]} ```