X posts

Our live demo of Jev doing x402 tool selection on a curated set of endpoints as requested @kleffew94. Worked better than we expected! https://t.co/dMxwIGedxN

Stop asking LLMs to “improve this tweet”. They're suck at it. But not JEV. I scored previous viral tweets, then made Jev check drafts against 88 criteria. 0.92 seconds. $0.00016 per check. It’s a simple loop anyone can reuse: Check the draft → get score → fix → check again Stop when the score stops improving. It doesn’t promise a viral tweet. But it shows what to edit next. Reply with the tweet yo

slackの自分の投稿を分類して、砂時計のように積んでいくアート作品。 どんな種類の投稿が多かったか、いろんな分類で比べてみれる。 Jevでの試作。 今の時代メディアアート作るの楽しくて良いな、１日目から表現に集中できちゃう。 (データは公開用に加工したものです) https://t.co/U81GsE285C

真面目なので。「JevはBERTでいい」というのを見てJevで作った中からBERTにしやすそうなやつをBERT版で作ってみました。 これは...非エンジニア門前払いなやつですね。結論、BERTでJev並みの分類器作れる人は天才。だからJevで驚くのはリトマス試験紙と言われてるのか... コスト面はBERTはAIコストかからないぶん判断基準の作成/更新のコストが大きくそこに１時間とかかけたらJevのコスト超えちゃうだろうから、 Jevというか生成AIのデメリットの「判断基準がブラックボックス化する」ことを嫌うシステムは分類基準を自分で作るBERTを使うべき、ということなのかな？と思いました。 こういう「それJevの必要ある？」という議論はすごく勉強になる。エンジニアさんたちはこういう議論を繰り返してベストプラクティス選定・使い分けを洗練させてきたのだろうなとすこし想像できて、とても勉強になり

“추석에 가족이랑 어디 놀러 가지?”를 카페 글 27만 건에 검색해봤다. Jev 예제를 맥에서 돌리다가, 직접 모아둔 한국어 카페 DB를 연결했다. 댓글도 58만 개 꺼내 검색에 넣었다. 일본 첫 여행, 교토 호텔, 분당 재건축, 마이크론 실적까지 질문을 연달아 바꿔봤다. 오사카와 후쿠오카를 비교한 댓글이 위로 올라오고, 재건축 검색에서는 실거주를 고민하는 글의 순위가 바뀌었다. 로컬에서 후보 20개를 찾고 Jev가 다시 정렬한다. 별도로 측정한 8개 질문의 Jev 응답 중앙값은 약 0.70초. 번역 없이 한국어 그대로 넣었다. 수집해둔 글을 검색하는 실험이라 최신 정보나 정답을 보장하진 않는다. 그래도 제목에 묻혀 있던 댓글을 다시 찾는 용도로는 꽤 재미있다.

Built a Chrome extension that uses Jev to decide in real time which posts in my X feed actually deserve to show up. It only lets through genuine content, on topics I care about, that's worth replying to. Everything else gets hidden before I ever see it. Runs entirely locally in the browser. Would you use something like this? Might open source it.

Humble beginnings using Jev. Trying to integrate it into Codex and help with speeding up browser use. Been working on it all day and running hundreds of trails. So far I've spent... $0.00725 https://t.co/yX3mPB41pL

Built for my mum, who has MND / ALS , using Jev @CompleteSkeptic @typesafeai thank you! AAC kit is exhausting - forcing a hunt through grids and having no context of what the other person says. I had Jev make a conversationally and context aware decision engine to pick… https://t.co/ESUAJcZ8r9

As promised… I just open-sourced the Jev X (twitter) Market Sentiment Analysis terminal powered by @typesafeai - Ingests 50 - 1,000 live tweets per crypto asset (e.g. BTC, SOL, ETH, etc.) - Scores each tweet for bullishness vs. bearishness using JEV - Combines social sentiment with live funding rates, RSI & volume - Detects short squeeze risks & outputs structured entry/stop/target cards - Built-in SQLite deduplication so you don't burn API credits on repeated searches Powered by Python, CCXT & TypeSafe AI's Jev model. Fork it, tweak it, build on it, or sell it ...but don't forget to enjoy

Built an AI dispatcher for Zammad tickets using Jev (TypeSafe AI). New ticket comes in, Jev decides which team owns it with a confidence score attached. Not a chatbot, a structured decision. Bridge is a tiny Ruby/Sinatra service. Open source, MIT. https://t.co/bXdJpGWHuw #jev

let's start simple. using jev for paragraph-ization: give a block of text and it will add paragraph breaks. I demo it using a text-only youtube client named jevtube. no videos, no distraction, just gets the transcript and adds paragraph breaks. https://t.co/rWuXnxBPUx

started using jev as a judge in @_moodshelf_ and the rankings in semantic search results and items classified by moods, have significantly improved, that too at much lower latency and cost

🔮 Built a real-time click prediction right over the DOM using JEV that feels like a crystal ball. Powered by jev-latest to predict the next clickable element and intent in 345ms per check, highlighting candidate elements on screen. It's extremely cheap to use and helps predict user intention, hitting 83% accuracy in testing. Usages that come to my mind: - Web performance optimization via predictive resource load - UX Research and testing - Progressive UI loading based on intention - Accessibility for reduced motor precision (highlight based on intention allowing to click, for example) - An

Built an SLO-aware inference router using Jev It uses Jev as a typed decision model to select the optimal LLM for each request based on predicted quality, latency, cost, and live backend load Releasing full walkthrough video soon https://t.co/XUjKj9Ryyn

Built jevmetrics — an OTel Collector component using Jev inference to score metric relevance, redundancy, and keep/reduce/drop decisions. Fail-open, with a deterministic policy still in control. https://t.co/1zLhyNjoEp #OpenTelemetry #Observability #Jev #Typesafe

I just had to explore using Jev with my existing Japanese print metadata extraction pipeline (where I use gpt-5.6 luna). Turns out that in some cases I could replace luna completely and save a bunch of money - in others I could augment what I had for higher quality! https://t.co/iSFOGPJjf8

.@jonptaylor recorded a detailed walkthough of Jev vs GPT-5.6 Luna as the "operator" element of a Pipecat speech interface pipeline. GPT-5.6 Luna: - 81.3% command accuracy - 1,008 ms median latency Jev - 92.6% command accuracy - 296 ms median latency A few notes here ... 1) We expected to see a big latency benefit. But the higher accuracy is maybe more interesting. Jev (with a bunch of code wrapped around it) is much better at turning messy transcriptions from a noisy speech environment into correct command structs. The very hand-wavy explanation here is something like: LLM too

Can we estimate a book’s literary quality and how enjoyable it is to read by using Jev to perform a multicriteria classification of its prose? I wanted to investigate that, so I built Salomón, a tool designed to do exactly this. I analyzed 32 books blind using Jev, and this is the map I got. Infographics, details and links in the thread.

Opencode sends every MCP tool schema to the model on every step. With 18 common MCP servers, that's ~90,000 tokens before the model even reads your request. I built a plugin using Jev that cuts it to ~7,300. (-92%) 🧵

Built a chrome extension that covers distracting youTube videos using Jev https://t.co/5TVn6cnLjD

I made an AI agent using Jev to reply to my wife We’re getting a divorce now and she wants to know who Jason is

I have a script running that's going through all of my magi-code sessions, extracting every bash command and then categorizing them by similarity using Jev. It's going to end up being about ~$0.60-$0.70 to do this for 4200 bash commands and take ~10 minutes with a very inefficient python script My goal here is to identify instances where the model keeps repeatedly writing python or bash scripts to perform very similar actions - and provide the agent with a small set of scripts that do those things for it so it doesn't keep wasting output tokens on writing repeated code.

Using Jev to catch Codex workers ignoring AGENTS.md. Combine live Codex output, Git changes, and AGENTS.md into one observation Use Jev to score the probability that the worker is drifting from AGENTS.md Steer the active Codex turn when drift crosses a threshold, stopping it if the drift continues Added to Foreman, which automatically watches and steers workers on the software factory floor. https://t.co/g9prp3tsy8

I swear this is the last Jev demo I'm doing... 🙏🏻 I'm using Jev by @typesafeai to get a third person opinion on my vibe coded projects... 👀 Checkout FirstScreen 🔗 https://t.co/nZnEsgxl1n As usual, it's again open source so feel free to star the repo if you want : ) For the past three days, I've had access to Jev, and I'm having so much fun playing around with it 🔥 I built Diffjury and JevArena in the past 2 days and shared with you guys on here, and today I built FirstScreen, where Jev basically takes a look at the website and quickly gives a verdict whether it's ready to ship or need

everyone's using jev to route between models. we put it on inbound sales replies the question isn't which model handles this turn, it's whether "sure" means yes. ours used to log that as a rejection and bin the lead [N] real replies through ai gateway. [x]ms p50, whole batch cost [z] papercut on the way: experimental_evaluate doesn't exist in ai v6, evaluation modality is v7 only, and nothing on the gateway model page says so. @vercel worth a line

i just shipped a small repo that will help you understand it perfectly.its a Real-time support triage + response bot using Jev and cerebras https://t.co/l4gFvr0GdW PS if you like it,approve my entry to the hackerhouse. I applied with rewantgoenka87076@gmail.com

exactly!! just shipped a real-time support triage + response bot using Jev and cerebras. https://t.co/l4gFvr0GdW

youre right. For example in a support ticket jev can classify way faster that which agent should get the ticket. I built a Real-time support triage + response bot using Jev and cerebras https://t.co/l4gFvr0GdW

I made a little experiment of using Jev to autonomously move around rooms of my 3d game. Headless operating based on textual position information. Rendering a screenshot at every step to be made into a gif. https://t.co/bCv0aoNxa9

just shipped a Real-time support triage + response bot using Jev and cerebras: https://t.co/l4gFvr0GdW this should help you get started!

I just built a Real-time support triage + response bot using Jev (from @typesafeai) and @cerebras Jev routes the ticket to a specialist agent (general / account / billing / technical) and decides whether a human should take it instead ,Cerebras then drafts the reply using whichever agent Jev picked. Repo: https://t.co/p6y4IQ2VhZ @callmeshuklaji @communidiyi @hackgoofer @dgo_almeida @sydneyrunkle

Built Jev Snake using jev from @typesafeai This is not your classic snake game fyi. 4 modes in total for this game -> Terrain Mode - You can generate objects REALTIME by writing ANY word. - The properties of the object will by given by Jev. - Your snake will react likewise with the object. HOT/COLD Mode - There are 70 fixed words that randomly appear. - Jev decides its effects - Far better you can type out word realtime. Jev will decide its effect too. WORD SNAKE Mode - There are 50 fixed words in vocabulary in our code. - Jev checks whether consequent words are correct or not - Want m

Jev join the race 🔥 Jev read the top 250 Hacker News stories 250 stories 2000 typed judgments in 16 seconds cost: $0.0095 about AI 32% Negative Tone: 41 ( 16% ) Built with Jev by @typesafeai https://t.co/RY3GiLTks9

Jev can tell if your resume is good enough for a job. I added a simple resume scorer using Jev by @typesafeai 1. You give your resume (Doesn't get saved) 2. PDF to Markdown conversion 3. Jev, does the analysis, across on 5 Dimensions 4. Your resume score Jev is quite fast btw. but he is very reasonable, so if the score is low, it is what it is.

#jev #typesafe #flutter #mobile #test Using Jev from https://t.co/wVCGdOU10g, I created a flutter package that allows you to write integration tests for Flutter apps using natural language. https://t.co/hG0mtW2FGp https://t.co/OkXWEkObbJ

Episode 285: Bendcoder We build an experimental coding agent from scratch using Jev from @typesafeai and new programming language Bend2 from @VictorTaelin. ...while playing WoW Forever on CoderOS; demoing the CoderOS GPU-accelerated multiplex panes and hand tracking (goodbye Hyprland, hello Jarvis); and introducing CoderQuest, a forthcoming game for commanding and upgrading coding agents with real-world prizes. Meanwhile in Azeroth, Eetum arrives in Thunder Bluff and purchases a guild charter. Bendcoder repo: https://t.co/UjtjgTEAZS

We did a test using Jev 1.13.0 on official LoCoMo data benchmark. Jev makes AtMem better at putting the right memory first, but it does not yet help AtMem find memories it missed entirely. The significant increase is on how often the correct memory was the very first result. Improved from 33.99% to 54.23% AtMem used about 1,593,522 token and costed about $0.0583 Jev remains as option in the next releases since AtMem Position is local first. We continue to explore more use cases for Jev

What I’m doing with #Jev from @typesafeai: Making Ergod, my multiplayer AI agent, better at the work between “let me check” and actually getting it done. Better research. Clearer comparisons. Fewer dead ends. Small decisions. Bigger capabilities. 🧵 A result can match your keywords and still miss your question. Jev helps assess what’s relevant. Alongside better page reading, Ergod can focus on useful evidence in long articles and wiki pages - with headings and tables kept in context. The follow-up is where this gets interesting. “What about the exceptions?” “Where did that figure come f

The smartass clipboard chrome extension I built with @typesafeai Jev is now open source. It reads the text you paste, breaks it into blocks and scans the form to understand what each field is. Jev then matches the right block to each field, verifies it against the original text and pastes only confident matches. https://t.co/0BfkIkQch9

just ran 1000 leads from https://t.co/ZngzUWDseG through jev gave it 10 questions per lead to check whether they actually fit our client 1,000 checks. about $0.006. each batch of 10 came back in under 2 seconds. less than a cent to qualify a real lead list? then i used @tryadlicio to pull what those brands’ customers were saying, jev helped pick possible outreach angles from the comments “congrats