X posts

i made an llm from first principles with Jev 29 yes/no questions per character: should the next key be a–z, space, comma, or period? highest probability gets append to it, then fed the updated text back in & repeat an autoregressive loop made out of a classifier https://t.co/OdiVS5MmM1

I got early access to @typesafeai's new Jev model and built a demo of it playing tetris The generation speed is so fast that it's pushing blocks down This feels like another "this changes everything" moment https://t.co/ND9rVcsPiD

Jev solved local harness/model routing I use a combination of Claude Code, Codex and Opencode as my local agentic stack and routing to other harnesses was always enforced in the system prompt/rules With a deterministic hook that Claude Code can decide before delegation, Jev helps to route to the right harness/model based on the task, and it's pretty accurate based on the intensity/intelligence of

We have jev at home. The arch is good for RL with state-dependent actionsets, so here is the same arch playing doom and chess (shittily) with two different controllers. https://t.co/OdhBY0jF9U

@typesafeai Just enabled true realtime UI generation, look at the speeds it can decide which components to build with just by passing it some JSON and wiring up to @shadcn Here we are starting off easy passing it pre defined options like a data table or donut chard to pick which one is best to display the data from

Ranked 2,500 jobs against a resume in ~60 seconds using @typesafeai's new model Jev It used ~13.5M input tokens which only costs 57¢ 16 recruiter questions per job. Skills, level, auth, interview odds. Each one comes back as a probability, then the ranking is just code. https://t.co/8MQvtVZvfP

“Movie night, but the baby is sleeping.” One sentence → six room settings. I built this demo with Jev from @typesafeai @CompleteSkeptic It chooses the settings; the app brings them to life. Real API responses, token usage, and estimated cost included. Try the room yourself 👇 https://t.co/EODjgQkuRk Describe a situation, press the arrow, and see what Jev chooses. What would you ask it?

JEV CAN DRIVE! 🤯 I hooked up Jev to the raw controls of a driving simulator, this video shows him driving in realtime. Realtime is the kicker here. Asking a model what to do every 50ms then rendering that into a "realtime" video is one thing, but in this simulation I wanted to give Jev the challenge of actually controlling a moving vehicle in a simulation that will never pause while he thinks.

you can turn Jev auto-regressive using a 254 token library this uses a derivative of the original GPT-2 set (removed things like "ing" since Jev had trouble generating full words and just kept repeating itself) it's quite limited in tokens albeit so don't mind the incoherence https://t.co/NqJPG1kyTA

got access to jev and instead of doing anything useful i made it play itself at 3D chess same model on both sides, slowly developing beef with itself you can watch jev vs jev or jump in and ruin its day https://t.co/VzDwchWqsI https://t.co/ly8d31pjSr

@typesafeai I think the main benefit of TypesafeAI is in enterprise, I created this demo in around 20 mins (it can be improved) I have been messing around with Typesafe for over 1 hour and used 1.5 million totals and it costed around 5 cents. https://t.co/u0Jd4bNy5F

I built a trading bot with Jev! Jev decides if it should "buy" or "sell", given the price feed of an asset pair, and executes real trades. It uses Monad to place the orders on Kuru's on-chain order book in every 300ms block. Demo link → https://t.co/vwl2SUu4jm https://t.co/Sda1G5tXKI

i gave @typesafeai jev messy bank transactions and let it clean up my ynab it normalizes payees, categorizes each transaction, asks me about the uncertain ones, and writes the results back https://t.co/c6F6ldIVQe