Skip to content

Case wall

Jev AI Guides cases

58 public posts filed under Guides. English summaries on this page, with a link to each case.

Last updated

These are the public posts filed under the Guides category.

Filter

24

quick tip we learned the hard way @ rivet: jev does not reason ...which is by design ...and we should have rtfm first ~ we started by prompt-engineer to ask som…

Guides

Nathan Flurry 🔩 @NathanFlurry

EN

@NathanFlurry shares a tip learned the hard way at Rivet: Jev does not reason, which is by design, and they should have read the manual first. A single prompt that asked for the severity of an issue, with SEV1 through SEV4 as outputs, gave meh results and defeated what makes Jev useful, because an LLM would spend reasoning tokens to reach a conclusion. After reading the manual they split it into small questions: did production go down or become unusable, are they blocked with no workaround, and is something in Rivet broken or failing. Code then returns SEV1 if blocked, SEV2 if down, SEV3 if broken, and otherwise SEV4. He says that works. When a ticket is weighted badly, they give it to an LLM with what they expected, the LLM adds it to a test suite and adjusts the decision logic, and they backtest the new logic on old issues. Within a few tickets the classifier was solid. He calls this a feature of Jev, not a limitation: more control over how each decision is weighed than a generic LLM, and easier to reason about, monitor, and backtest how agents will behave. In their internal support tool the Jev question counts are 13 for support classification and customer status, 6 for severity, 6 for team replies and problem identification, 6 for follow-up timing, 4 for linked dependencies and later conversations, 2 for the addressed teammate, and 2 for conversation grouping. He ends with: Jev is awesome. A screenshot of a real-world classifier is attached.

quick tip we learned the hard way @ rivet: jev does not reason ...which is by design ...and we should have rtfm first ~ we started by prompt-engineer to ask som…

Poster from Annatar.md (@AnnatarXBT)

Guides

Annatar.md @AnnatarXBT

EN

@AnnatarXBT posts about a 14-page paper on building a company brain: a cheap decision layer in front of an expensive reasoning model. The architecture is pulled from the TypeSafe and Anthropic docs. Opus 5.5 thinks, Jev decides, and the code holds the branch. You ask a typed question and get a typed answer. Jev returns it with a calibrated probability in 0.44s for $0.00035. A 0.999 goes straight into the if statement, and about 99% of turns end right there. The same gate, placed next to Fable 5.2, is broken down in the article below. The post quotes an earlier @AnnatarXBT post, which calls itself a full harness guide for Claude Fable 5.2 and Jev engineering: one evening with it and Claude does the work of a small dev team, and no $550/month course is needed. That quoted post links https://x.com/i/article/2120740198405357568.

a 14-page paper on building a company brain: a cheap decision layer in front of an expensive reasoning model the architecture, pulled from the TypeSafe and Anth…

Quoted @AnnatarXBT

Poster from Jordan (@nolansym)

Guides

Jordan @nolansym

EN

@nolansym posts a new AI pattern, Jev Composite Scorecard. Ask atomic score questions, then weight them in code. The demo weights are clarity 40%, completeness 40%, and risk 20%, on a PR description. He built it with @aisdk v7 and the @shadcn registry.

New AI Pattern - Jev Composite Scorecard 🤝 Ask atomic score questions, then weight them in code. ✅ Clarity 40% · completeness 40% · risk 20% ✅ PR description de…

Coding Agent 最浪费 Token 的地方,可能根本不是写代码。 最近读了一份《Jev Engineering for Coding Agents》,整理了 TypeSafe 创始人 Diogo Almeida 对 Agent 架构的思考。 其中几个观点值得关注: 写代码只占约 4%~10% 的 Token,…

Guides

旺旺 @TiagerBao

ZH

@TiagerBao posts Chinese notes on a third-party compilation, "Jev Engineering for Coding Agents," of TypeSafe founder Diogo Almeida's thinking. He writes that writing code is only about 4–10% of tokens, and that the notes are not a shipped complete plan.

Coding Agent 最浪费 Token 的地方,可能根本不是写代码。 最近读了一份《Jev Engineering for Coding Agents》,整理了 TypeSafe 创始人 Diogo Almeida 对 Agent 架构的思考。 其中几个观点值得关注: 写代码只占约 4%~10% 的 Token,…

Poster from Akshay 🚀 (@akshay_pachaar)

Guides

Akshay 🚀 @akshay_pachaar

EN

@akshay_pachaar writes that agent memory pipelines store too much, and that a decision model like Jev is useful when the app has its own rules for what may enter long-term memory.

Your agent remembers too much. Decision models like Jev can fix that. Let me explain: A common agent-memory pipeline sends conversations, tool results, and task…

Quoted @akshay_pachaar

Poster from Mr. Buzzoni (@Mr_Buzzoni)

Guides

Mr. Buzzoni @Mr_Buzzoni

EN

@Mr_Buzzoni packs a Jev plus Opus 5.5 company brain into a 14-page PDF taken from the TypeSafe and Anthropic docs, and lists ten steps from a typed Jev answer through to judging every run.

JEV + OPUS 5.5 IS INSANE FOR BUILDING A COMPANY BRAIN I pulled the whole architecture out of the TypeSafe and Anthropic docs and packed it into a 14-page PDF th…

Poster from Gul Saeed Khattak (@Ldataengineer)

Guides

Gul Saeed Khattak @Ldataengineer

EN

@Ldataengineer walks through OntoSmith, an open-source knowledge-graph builder where a cheap model proposes facts, plain code checks the quotes, and TypeSafe's Jev, as a separate judge, decides what gets written.

OntoSmith: 𝗧𝘂𝗿𝗻𝗶𝗻𝗴 𝗬𝗼𝘂𝗿 𝗔𝗜-𝗕𝘂𝗶𝗹𝘁 𝗞𝗻𝗼𝘄𝗹𝗲𝗱𝗴𝗲 𝗚𝗿𝗮𝗽𝗵 𝗙𝗿𝗼𝗺 𝗮 𝗚𝘂𝗲𝘀𝘀𝗶𝗻𝗴 𝗚𝗮𝗺𝗲 𝗜𝗻𝘁𝗼 𝗣𝗿𝗼𝗼𝗳. You point an LLM at a pile of documents and ask for a knowledge graph. It give…

Poster from Archimedes (@ArchimedesLore)

Guides

Archimedes @ArchimedesLore

EN

@ArchimedesLore walks through a typed Jev pipeline: STATE cards, seven typed questions in one shot, parallel choice/score/noul at a 256 ms median with zero tokens out, and a 0.85 gate before code acts.

WHOEVER LEAKED THIS JEV DECISION PIPELINE SHOULD BE WRITING THE RFC I thought routing an agent through five steps meant five prompts, five parsers, and five pla…

Quoted @0xCodila

What a fast, cheap and reliable judgment model has in common with humans and where it differs. Jev from @TypeSafeAI can support useful cognitive tests at a scal…

Guides

Alexander Pico @xanderpico

EN

@xanderpico writes that a fast, cheap and reliable judgment model shares some traits with humans and differs in others, and that Jev from @TypeSafeAI can support cognitive tests at a scale that used to be impractical.

What a fast, cheap and reliable judgment model has in common with humans and where it differs. Jev from @TypeSafeAI can support useful cognitive tests at a scal…

One thing that can be tricky to figure out is how to use Jev's primitives to handle more than one intent in a single user query. My favorite way to approach thi…

Guides

Allie the Icon @allietheicon

EN

@allietheicon describes a "shopping cart" pattern for more than one intent in a query: recursively pick the next thing until Jev says done. She used it for a Slack coffee bot (DoorDash CLI) and a smart-home handler.

One thing that can be tricky to figure out is how to use Jev's primitives to handle more than one intent in a single user query. My favorite way to approach thi…

Poster from Jordan (@nolansym)

Guides

Jordan @nolansym

EN

@nolansym shows a "Jev JSON Render" pattern: English in, UI out. It picks prepared fields and never writes new copy, and the preview can be edited after it composes. Built with AI SDK v7 and the shadcn registry.

New AI Pattern - Jev JSON Render 🤝 English in, UI out. ✅ Picks prepared fields, never writes new copy ✅ Edit the preview after it composes built with: @aisdk v7…

Jev can return a category even when none of your categories fits. It can return a bounding box even when the target is absent. That's an important API contract…

Guides

Milhous @milhouscc

EN

@milhouscc explains that Jev can return a Choice even when no category fits and a Box even when the target is absent. He recommends companion Noul/presence checks, treating confidence as distribution concentration not correctness, and keeping application policy separate from model output. Design reading of the docs, not a deployed benchmark.

Jev can return a category even when none of your categories fits. It can return a bounding box even when the target is absent. That's an important API contract…

Poster from Varian (@0xVarian)

Guides

Varian @0xVarian

EN

Jev router writes each handoff as queue/[destination]/[uuid].json. CHOICE names the next worker; below 0.85 goes to review.

WHOEVER LEAKED THIS JEV ROUTER CODE SHOULD BE REPLACING KAFKA I thought handing a job between workers meant Kafka, Redis, a schema registry, and a retry topic t…

An ant colony in your browser: every ant asks a decision model where to go – and you, the queen, steer it with one sentence. 🐜👑 A tutorial in 8 steps with Pytho…

Guides

Gerhard Völkl 🐘 @gkvoelkl@sueden.social @gkvoelkl

EN

@gkvoelkl shows a browser ant colony: every ant asks a decision model where to go, and the queen steers it with one sentence. An 8-step Python and Streamlit tutorial in English and German.

An ant colony in your browser: every ant asks a decision model where to go – and you, the queen, steer it with one sentence. 🐜👑 A tutorial in 8 steps with Pytho…

Poster from ZazenCodes (@ZazenCodes)

Guides

ZazenCodes @ZazenCodes

EN

@ZazenCodes walks through Jev 1.14 images, Box questions, 262K context, and a screen-watcher app.

Jev can see 👀 0:00 What's new in Jev 1.14 2:39 Demo 1: Analyze Images with Jev 11:37 Demo 2: Draw Boxes Around Objects 16:26 Demo 3: Analyze a Long SEC Filing 2…

Poster from ericosiu (@ericosiu)

Guides

ericosiu @ericosiu

EN

Single Jev prompt reviewed 3,208 pages in ~3 minutes and labeled fix, merge, or refresh—shareable Claude Code recipe.

Steal this. Our @typesafeai Jev prompt checked all 3,208 pages on our website in 3 minutes and told us which ones to fix, merge or refresh. Drop it into Claude…

Is the cat plotting revenge? 😼 We moved a classifier off an LLM judge and onto Jev, TypeSafe's decision model. It matched the LLM on accuracy and runs much fast…

Guides

Amy Slawson @amzslaw

EN

@amzslaw moved a multi-label classifier from an LLM judge onto TypeSafe Jev; after question and cutoff tuning it matched LLM accuracy and ran faster.

Is the cat plotting revenge? 😼 We moved a classifier off an LLM judge and onto Jev, TypeSafe's decision model. It matched the LLM on accuracy and runs much fast…

Inspired by "Jev", an AI model specialized in decision-making, a number of downloadable open-weight Jev models have been emerging. Today, I will pick five Open-…

Guides

Toshifumi Kuga @ToshifumiKuga

EN

@ToshifumiKuga article: five Ollama Open-Jev models vs original Jev on bank-complaint product classification.

Inspired by "Jev", an AI model specialized in decision-making, a number of downloadable open-weight Jev models have been emerging. Today, I will pick five Open-…

Build note from JobPly: "What's your notice period?" with options like Immediately, 2 weeks, 1 month looks like a text question. It isn't. It's a pick-one. So i…

Guides

Hasnain Bharmal @BharmalHasnain

EN

JobPly maps “notice period” options to a Jev pick-one so the answer is one of Immediately / 2 weeks / 1 month.

Build note from JobPly: "What's your notice period?" with options like Immediately, 2 weeks, 1 month looks like a text question. It isn't. It's a pick-one. So i…

Poster from Archimedes (@ArchimedesLore)

Guides

Archimedes @ArchimedesLore

EN

@ArchimedesLore describes a Jev decision pipeline: STATE, typed questions, Jev System One, a 0.85 gate, then deterministic code.

WHOEVER LEAKED THIS JEV DECISION PIPELINE SHOULD BE WRITING THE RFC I thought routing an agent through five steps meant five prompts, five parsers, and five pla…

Quoted @0xCodila

Poster from ericosiu (@ericosiu)

Guides

ericosiu @ericosiu

EN

@ericosiu argues Jev Images is underused for marketing: AI search visibility, ad creative gaps, CRO audits, pain-point reports.

Jev Images is a BIG deal for marketing, and it's being slept on. Scroll down for an unfair advantage. Some use cases: • AI search visibility report • Competitor…

Poster from Anshumani Ruddra (@baboonzero)

Guides

Anshumani Ruddra @baboonzero

EN

@baboonzero uses Jev, TypeSafe's System 1 model, to decide where a teaching explanation starts, ends, and stands alone, in a Super IC write-up.

I have more than a hundred hours of recorded teaching. Finding useful two-minute clips means deciding where an explanation starts, where it ends and whether it…

how to add a decision model to your AI workflow 1,000 yes/no decisions can cost around: ~$10 on GPT-6 Astra ~$4 on Claude Opus 5.5 ~$0.04 on JEV use language mo…

Guides

Shann³ @shannholmberg

EN

@shannholmberg's six-step guide: use LLMs for planning/coding and JEV for yes/no decisions (~$0.04 per 1,000 vs dollars on GPT/Claude).

how to add a decision model to your AI workflow 1,000 yes/no decisions can cost around: ~$10 on GPT-6 Astra ~$4 on Claude Opus 5.5 ~$0.04 on JEV use language mo…

Unofficial developer resource. Not affiliated with TypeSafe AI. These posts belong to their authors. jevnotes.com does not host the videos.