we gave jev 12000 synthetic records of Diabetes it was able to classify then in seconds jev is a classifier and it works insanely well when paired with an LLM which acts a rulebook creator https://t.co/qX9lGuOFBU
Did a Computer Use app for mac using Jev @typesafeai and the speed is insane, note this is still very raw yet, but honestly I'm truly impressed with this experiment and there's so much so can create and improve. Next will be experimenting this with @avocadoai_co super agent and will post later the progress
I wanted to see where Jev might fit into an accounting workflow, so I set up a test using 250 synthetic bank transactions. Each model got the same transaction details, a simplified chart of accounts, some bookkeeping rules and a written summary of the evidence. The job was to pick a category and decide whether the transaction needed an accountant to review it. So it was making decisions from infor
okay but why stop at if statements went one level lower and put Jev in an instruction set. CHOOSE, SCORE, TEST, JUDGE now a register can hold “97% chance that was kind” and the next instruction has to deal with it built a tiny computer to try it ⬇️ https://t.co/EE8yYqwuHc
tested a homebrew Jev reflex gate locally on a 4GB GTX 1650 Ti: the premise: instead of burning slow thinking tokens on standard developer collisions, use a small 2B model as a sub-20ms System 1 decision gate. base 2B models hit an 83.3% catastrophic action rate (e.g. 57.0% probability of reformatting disk on a port 8080 collision). trained an 8.6MB LoRA adapter (4-bit NF4) into a JevMiniCPM refle
Jev doesn't write, but it can judge... FAST! I made a first-person puzzle game powered by Jev. You're 8. You ate every cookie. Mum's home in 15 minutes. Hide the evidence. Make your choices. Let the probabilities decide if she buys it. Play: https://t.co/0ciVHr3SXchttps://t.co/8mMbyR6KsR
What happens if you put #Jev inside a game loop? I built a poker roguelike and gave it full control. It can buy upgrades, discard cards, choose hands and play the entire run by itself. No chatbot. No scripted NPC. Jev is literally playing the game. Built this as an experiment in using fast decision models for real-time game agents. 🎮 https://t.co/jX7wRXYjfG 💻 https://t.co/cFYvWRrDUR Watch the righ
I used Jev to create a platform for A/B testing tweets, LinkedIn posts, and YouTube hooks across hyper-specific demographics. The best part is that it simulates responses from real people rather than imagined personas. It turns Jev into a testing ground for how specific audiences might respond before you publish. Link below.
like @levelsio I often suffer from picking good emojis for any label. made a little jev app that decides from all emojis what fits best and goes super fast! https://t.co/fEAAotrRMM
JEV PLAYING BULLET CHESS ON ANDROID - 316 ms median decision latency - 1.35 s median observation to tap execution - $0.000052 estimated cost per decision - 110 seconds total runtime - $0.00218 estimated total cost wtffff https://t.co/O8sgnwXU90
1681X posts·Games & real timecost$0.000052 estimated cost per decision; $0.00218 estimated total costtime316 ms median decision latency; 1.35 s median observation to tap executionOriginal source ↗
Claude Code changed an auth check to return true. Tests passed. It was ready to finish. I built jev-preflight: @typesafeai‘s Jev flags the risk and sends Claude back once. 8 risk axes · 1 request · max 1 re-check Open-source beta ↓ https://t.co/WdeIiJ8O6hhttps://t.co/FYoS7Ukw4H
Jev as a docs confusion heatmap checker! 🔍🔥 Did a small experiment with Jev by @typesafeai. We fetch a technical docs site and ask Jev to judge how difficult each section might be for a beginner. It looks for: - unexplained terminology - assumed prior knowledge - unclear conceptual relationships - ambiguous instructions - excessive information density Then we turn those judgments into a confusion
I gave @typesafeai's Jev Flappy Bird to play 😂 Across 34 minutes of runs: 1,264 pipes passed, 3 deaths, with the game ramping up to 2.5x speed. All it does is pick flap or wait every ~280ms. No reasoning, and it costs about $0.26/hour. Kinda wild how much you can do with decisions this fast and cheap
i got Jev inside a clinical workflow. this note mentions 6 conditions. OpenMed reads each span in context, Jev makes 6 typed decisions, and code lets 1 into the current problem list, blocks 4, and sends 1 to human review. https://t.co/ZeP59oZMpx
Vibe coded a 2D combat arena with Jev. The goal was to experiment with building an autonomous boss AI that actually fights smart and tries to win in real time. It’s not quite unbeatable yet, but watching it self-play with different personality vectors has been fun. Toggling state being sent and prompts was changing the strats quite a lot. Would love to hear any ideas or prompt suggestions to make
I wired TypeSafe AI’s Jev directly into a live crypto market. It watches BTC, ETH, and SOL and makes typed BUY / SELL / HOLD decisions every few hundred milliseconds. No chat. No long reasoning trace. Just: live market state → Jev → decision → action Here’s what it looks like in real time ↓
I let JEV qualify 3,000 companies in milliseconds Scrape, enrich, ICP score, verify, CRM write, draft. JEV decided what runs next. 1,260 came out as fit. 61 it wasn't sure about, those went to a human for review. 41 seconds. $0.008. Agencies could be saving $150/m on heavy qualifications with their big target lists Live in Orbit for any GTM motion. Comment "JEV" and i'l send the link so you can tr
JEV is king at categorization! 🚨 JEV beats the TOP embedder (Qwen3-Embedding-4B) at categorization! 46% vs 97% accuracy! Not a simple win! We benchmarked +1000 TODO titles: - JEV scored 97% accuracy compared to Opus 5 reference. Qwen3 only 46%! - Embedder costed 0.15$/million TODO vs JEV 10$/million TODO Extremly cheap and extreme accuracy while staying superfast! Hopw you guys love it too! Blogpo
I upgraded Jev’s Kitchen Chaos after V1. Biggest change: the 4 chefs no longer make decisions at the exact same moment. Requests are now staggered, so each chef can react independently. Then I reran Jev vs Claude Opus 4.8 vs GPT-5.6 Sol. Results: 🥇 Jev: 495 🥈 Claude Opus 4.8: 455 🥉 GPT-5.6 Sol: 450 Watch the run 👇
Been poking JEV (https://t.co/0CYgSpQWpG) in two modes. Games: 1) Shooting game, me vs JEV, same targets. JEV 14 hits, 0 misses, 100% accuracy. Me about 82%. Avg decision about 352ms. 2) Crate Gate (10 Deep Vault). Spatial puzzle with live judgments and confidence. Work: My research bot (Olvia) runs a lengthy government document application. JEV is the gate before each execute step: what next, eno