Video wird geladen...

Video konnte nicht geladen werden

Zur Startseite

How does Jev work? Jev takes two things: your input and an output schema. The schema gives the structure of the answer you want. I send an email. My schema asks for two fields: is spam, and a category from marketing, finance, or spam. Jev returns: - is spam:...

12,032 Aufrufe • vor 1 Monat •via X (Twitter)

3 Kommentare

Profilbild von Joshua Pi'Rwot
Joshua Pi'Rwotvor 1 Monat

Writing the schema is where most of the real work sits, and it is easy to underrate. Once you can state the shape of a correct answer field by field, you have already decided what counts as correct. The model filling it in is doing the easier half. When output quality is poor it is usually because the schema was vague, and the prompt gets blamed for it. It also makes failures easier to spot. A missing field shows up immediately, where a weak paragraph can sit there for weeks with everyone assuming it is fine.

Profilbild von Jay
Jayvor 1 Monat

the schema is doing the real work here

Profilbild von AI Mastery Guide
AI Mastery Guidevor 1 Monat

Finally someone explains Jev simply

Ähnliche Videos

Jev is quietly taking over the agent stack. If you already have a Jev API key and don't know what to build first, steal this checklist: 1. jev-opus Runs Claude Code on Opus 5.5 and asks Jev before every call how much effort the next step needs: low, medium or high. Code enforces the floors and ceilings, and the prompt cache stays intact. 2. jev-belay A Claude Code Stop hook that won't trust "done" without evidence. If files changed and nothing passed since, one Jev call checks the transcript before the agent is allowed to stop. 3. jev-skill-router On every prompt, Jev picks one skill from everything you have installed, or none. It starts in shadow mode and only logs, so you see its picks before you trust them. 4. codex-jev-router Jev picks the model and reasoning effort for each Codex subagent. When Jev is unsure, it falls back to Sol. 5. jev-commit A pre-commit hook. One Jev call checks whether your commit message actually matches the staged diff and flags debug leftovers and unmentioned work. 6. is-malicious Scans source, config, build and CI files with Jev and points at the suspicious lines before the code ever runs. 7. jegrep Semantic grep in Rust. Jev scores your files with probabilities, no embedding index and no background daemon. 8. jev-scout An MCP server that scores every search query, result and fetched page your agent touches for relevance and credibility, with a live decision dashboard. 9. hunch Probabilistic control flow for Ruby. if Hunch.likely?("fraudulent", given: order) branches on a typed Jev answer. 10. pytest-jev Write plain-English claims about your output, and the test passes only when Jev clears every claim above 0.8. Copy any of these Jev builds, then read the full Jev setup below ↓

Chen

29,260 Aufrufe • vor 1 Monat

People already built a browser agent, a Claude Code bridge and a LangGraph router on top of Jev. Here are all 10 repos. If you have API access and don't know where to start, save this: 1. jev-browser Real browser control through MCP, CLI or as a library. At every step Jev answers two things: which element to touch, and whether the goal is done or the run is stuck. > 2. jev-browser-agent Picks the operation and the target in a single request, around a second per step. Anything under 0.6 confidence goes up to a bigger model or back to you. > 3. typesafe-jev-bridge OpenAI-compatible bridge with zero dependencies. Typed yes/no, choice and score answers inside Claude Code, Cursor, Cline or any OpenAI SDK, via CLI or HTTP. > 4. pilot-typesafeai-jev Jev wired into LangGraph and Deep Agents. Three questions in one request, plain code chooses the route, and the chat model only spends tokens on the route that won. > 5. building-with-jev-skill Agent skill for writing code that calls Jev. How to phrase questions, structure state, set thresholds, and debug a question that keeps answering wrong. > 6. jev-skill Guided setup and eval workflows. Checks what's installed without changing anything, never swaps your provider behind your back, never makes up probabilities. > 7. awesome-jev (yibie) What people actually shipped with Jev, sourced and sorted by category, with the limitations written down for each entry. > 8. awesome-typesafe-jev SDKs, live demos, agent tools and independent evals. Shows where every vendor number comes from. > 9. awesome-jev-by-typesafe Use cases, patterns and starter code backed by evidence. The 70 to 500 ms latency figure is labeled vendor-reported, not measured. > 10. awesome-jev (cobanov) Biggest catalogue so far: SDKs for Go, Java and JS, MCP connectors, guardrail experiments and studies. > Copy these complete Jev blueprints - then read full Jev setup below ↓

Asteri

16,629 Aufrufe • vor 1 Monat

This is completely f*cking wild. A Stanford AI researchers gave Claude Code a second brain with JEV. The result? 80+ billion data points processed every 14 minutes. here's how it works: Jev stands at the door. every row has to get past it first. the easy ones it waves through on the spot. only the ones it can't call get sent upstairs to Claude. the old way: everything -> Claude -> big bill the new way: everything -> Jev -> the tricky few -> Claude Jev also tells you how sure it is on every call. that number decides who goes upstairs. just don't pick it out of thin air. test it first. the 20-minute setup: 1. take 200 real rows from your data 2. run them through Jev AND Claude 3. sort Jev's answers by how sure it was 4. find the point where Jev stops disagreeing with Claude. that's your cutoff 5. above it, Jev's answer is final. below it, the row goes to Claude then do the math. on a public benchmark Jev cost $0.015 per 1,000 rows and Claude cost $0.80. if Jev settles 80% of your rows: $0.015 + (20% × $0.80) = $0.175 per 1,000 rows instead of $0.80 that's ~78% off, with Claude still handling every hard case. 3 rules that keep it working: -> ask all your questions in one Jev call. topic, urgency and sentiment together, not three separate requests -> once a week, send 50 of Jev's "confident" answers to Claude anyway. if they start disagreeing, raise the cutoff -> never let Jev make the final call on anything you can't undo. refunds, bans, deletions go to Claude or a human where this pays off right away: -> support tickets: route by topic and urgency -> leads: hot, warm or skip -> comments: fine, spam or needs a look -> spreadsheets: fix category labels across thousands of rows Claude thinks. Jev picks. code ships.

kiosa

14,006 Aufrufe • vor 26 Tagen

this is f*cking gold 20 GitHub repos with 500K+ combined stars that will level up your JEV workflow AGENTS > jev-ultrafast: Jev picks every click and DOM target, a small LLM only types > hermes-jev-skills: routing, memory, compaction and skill picks in one pack > typesafe-computer-use: OCR reads your Mac screen, Jev picks the next click MEMORY > fast-jev-compaction: scores every tool call keep, truncate or drop instead of summarizing > jevmem: project memory for Claude Code, Cursor and Codex, updated every turn > jev-second-brain: your Obsidian vault, Jev judges which notes duplicate, revise or contradict SAFETY > jev-guard: a gate before every tool call, Jev scores the risk, you set allow, ask or deny TOOLS > skills: the official TypeSafe skill for Claude Code and Codex > system-one-adapter-python: dry-run your questions on an ordinary LLM before you burn a Jev key > jev-mcp: claim checks, screening and ranking as MCP tools > typesafe-mcp: plug Jev into any MCP client > json-render: Vercel's generative UI, where Jev picks the components OPEN MODELS > SemIf-OpenJev: semantic ifs from frozen open models kev: Jev-like models on Qwen that run on your MacBook > laya-mlx: the Laya decision engine on Apple silicon > clm: an open System One model with Choice, Noul and Score > jevlike: train your own Jev-like model TRADING > jev-trader: one buy or sell decision per Monad block START HERE > awesome-jev: the biggest map of everything built on Jev > awesome-jev-by-typesafe: use cases, patterns and starter code bookmark it before your next build

NO1ennn

18,445 Aufrufe • vor 1 Monat

Jev is cool. So is it's OSS companion, Laya. The Latest Cool Thing In AI™ tends to get a lot of hype, sometimes without everyone even understanding it. So... what is this thing? Jev is an AI model that consumes input and produces output VERY differently than chat, claude, grok. The input is two things: 1) Text state to assess. Email, html, code, whatever. 2) A set of questions which will be asked about the attached state. The canonical example from TypeSafe's docs is to identify the urgency of a support ticket. We pass the model the customer text + a single noul question "is this urgent?". Jev returns a full set of JSON. This JSON is not generated with token-by-token autoregression. Jev is not trained to produce sequences of text tokens, rather to answer questions, and guarantees well-formed responses. In the example below, we see it produces a 0.99 probability (on a 0-1.0 scale) that the answer is "yes." Jev supports exactly three types of questions (seconds example in video): a) Noul: 0–1 probability that the answer to a yes/no question is "yes." b) Choice: Ask question with pre-defined set of answers. Jev chooses the best and assigns probabilities to each. c) Score: Ask question with pre-defined scale of answers. Jev produces a position on the scale. Jev computes answers for all questions in parallel, making responses super fast even for many questions in a single request. This might seem like a narrow set of capabilities, but in the right contexts leads to incredible potential. It also makes for a useful API / primitive for programming, since the outputs are... *ahem*... type-safe and predictable in structure. Jev is not going to replace LLMs for writing your code, auto-generating your docs, or being at the core of an agent harness. But Jev IS incredibly cool, and will be used to build a lot of amazing tech. Hope this helps.

Ben Dicken

42,183 Aufrufe • vor 1 Monat