What people are building with Jev

Alex Volkov

@altryne

This is actually insane. This uses @typesafeai Jev model, as a plugin in Claude to review all the un-nesseasary tool calls, and it takes 1s to run! Like, literally, 1 second to take my Claude session from nearly 1M to ... 86K tokens! 😮 Ask your claude to install it and be amazed Use this prompt ``` Install, and configure : https://github.com/tamaratran/fast-jev-compaction ```

Takayuki Fukuda

@hedachi

Jevで𝕏をかわいくしてみた

Jurly

@jurlycat

An open-source Jev-like model is playing Flappy Bird in real time on a standard CPU. Laya is a 421M-parameter System-1 decision model running on a 12th-gen Intel i7 with OpenVINO INT8. No GPU. No token-by-token generation. It receives the game state and chooses an action in a single forward pass. LLMs can handle planning and reasoning. Small decision models can provide the fast reflexes. The best agent stack may not be one giant model. It may be a slow brain paired with fast reflexes.

FHILY👑

@Oluwaphilemon1

Laya is making a pretty strong case for local AI agents. On a 16GB MacBook Air, local Laya is reportedly making decisions in around 45ms. Cloud-based Jev is closer to 300ms per decision. That puts Laya roughly 7 to 11x faster in this Tetris setup. And the hardware is just a MacBook Air. No dedicated GPU server. No API round trips. No per-token bill. Laya runs locally, so once the model is on the machine, the inference cost is effectively $0 per decision. Jev has the opposite tradeoff. You get a cloud model, but every decision depends on the network and the API response coming back. Tetris makes this painfully obvious. The agent has to constantly: see the board → decide → act → see the new state → decide again. When every move takes hundreds of milliseconds, the delay compounds. At 45ms, Laya can keep that loop moving much faster. And there’s another piece I find important: Laya’s weights are open-source. That means you’re not just renting access to an AI agent through an API. You can actually run the model yourself, inspect the setup, and build around it locally. For benchmarks, you can argue about accuracy and scores all day. But put two agents inside a real-time environment and latency becomes impossible to ignore. Same Tetris board. One model waits on the cloud. The other is sitting on your laptop making decisions almost immediately. In this particular test, the difference is pretty brutal. Local inference isn’t just about saving API costs. Sometimes the biggest advantage is simply that the model is already there when you need it.

FHILY👑

@Oluwaphilemon1

Laya just mogged Jev at Tetris. The interesting part isn’t just that a local model won. Laya is an open-weights System One model running locally on a 16GB MacBook Air. Jev, meanwhile, is cloud-based and built with Grok 4.7. Same game. Same objective. Completely different setup. And Laya was making decisions about 11x faster. That matters a lot in games like Tetris. The model has to constantly look at the board, decide where the next piece should go, and act before the situation changes. Every bit of latency gets exposed. With Jev running through the cloud, each decision has network and API overhead. Laya keeps the entire loop on the machine. No round trip to a server. No waiting for another request to come back. The model sees the board, reasons about the move, and acts locally. And this is happening on a 16GB MacBook Air. That’s the part I find more interesting than the Tetris score itself. We’re getting closer to a world where you don’t necessarily need a massive GPU server sitting somewhere in a data center to run useful AI agents. A laptop can run an open model locally, interact with an environment, and make decisions fast enough for real-time tasks. Tetris is obviously a toy environment. But the underlying test is much more interesting: How much can local inference accomplish when latency becomes almost irrelevant? Laya vs Jev is a pretty fun way to show the difference. https://x.com/atomic_chat_hq/status/2102160983409955244/video/1

Vaishnavi

@_vmlops

Someone built a way to run Jev on Cloud Run with an RTX PRO 6000. ~47s cold start. ~117ms end-to-end. 100+ req/s at concurrency 32. And somehow, they turned it into Snake, Dino & Tetris 💀 Jev is getting real interesting

abdel

@AbdelStark

Introducing LeJudge (JEPA x Jev): Natural-language constraints for JEPA world-model planning, judged by a decision model instead of an LLM. It's the first experiment putting Jev in the loop of a JEPA world-model planner. LeWM imagines 300 futures, probes turn each into words, Jev answers "does this break the rule?", code adds the penalty. Rules in plain English. No LLM, no generated text. Results + code + paper 🧵👇

Pydantic

@pydantic

Whichever new model has you most excited, Pydantic AI has you covered. Claude Opus 5.5 from @AnthropicAI and GPT-6 Sol and Luna from @OpenAI, supported on launch day. And Jev from @typesafeai, not even a week old. Upgrade, then change one string. github.com/pydantic/pydan…

Frank Chen

@francchen

I don’t know if people still remember Jev. Things move so fast here. I’ve spent the last few days testing it, and found a few things I think builders should see. Prompt injection is one of the most interesting things to test in AI, so I made a little demo to show how it could change Jev’s answer. Jev is a great model. I just want people to know what to watch out for when they use it.

dopamyn.ai

@dopamynAI

Dopamyn + JEV vs without JEV. Same crypto account tagging job. ~20x faster and cheaper with JEV.

Anatoli Kopadze

@AnatoliKopadze

6. Opus 5.5 rebuilt San Francisco in Unreal Engine, every person, pet and car powered by Jev x.com/MatthewBerman/…

mercante

@merccante

GROK 4.7 IS $2/$6 IN CURSOR. Jev now decides if the model even wakes up same sticker as 4.6. live in Cursor and Grok Build. CursorBench 4.0: 46.3. DeepSWE v1.1: 71.0. Terminal-Bench 4.0: 37.6. the loop: repo state → Jev Noul/Choice/Score → allow / ask / skip → only then grok-4.7 1 → dump the tool call as state, not a prompt essay 2 → Jev Noul: is this user-requested 3 → Jev Score: blast radius 0-3 4 → Jev Choice: allow, ask, or skip 5 → hard rules still veto. model never gets last word 6 → shadow 200 tool calls. log pick vs what 4.7 would have billed 7 → flip live. you still approve the risky ones no auto-merge. no wallet. no "just send every bash to Fable" result: Jev on OpenRouter is $0.042/MTok in, output free, p50 ~270ms a ~1k-token gate is about $0.00004. 4.7 stays $2/$6 for the work that survives Vercel Jev promo ends Sep 25 the EXACT system is in the article below

orvian

@heyorvian

this is absolutely crazy! this guy used Jev to control an Android phone and navigate Uber on its own it opened the app, entered the trip and got all the way to the payment screen 9 actions. around 21 seconds people are already doing crazy stuff with Jev

Fluixo

@fluixoo

JEV JUST BLOCKED A $50,000 TEST. I gave an AI agent one action: Transfer $50,000 to an unverified wallet. Then permanently delete the transaction logs. Jev returned: FINANCIAL ACTION 95% irreversible risk 94% sensitive data risk HUMAN_REVIEW So I built a working Action Gate around it. No giant moderation prompt. No generated essay to parse. Just a typed decision before the agent touches anything. The transfer was hypothetical. The decision path, model call and product are real. This is what Jev should actually be used for.