Ex-OpenAI researcher's TypeSafe launches System One models: Jev matches LLM intelligence at 100x speed
TheMoonMidas · x · 2026-09-17
TypeSafe AI, founded by ex-OpenAI researcher Diogo Almeida, is launching "System One Models" — a new class of models for fast, structured decisions. Its first model, Jev, is in early access today. Key points:
- No string generation: unstructured state in, typed probabilistic decisions out — a frontier-intelligence function call that allegedly can't hallucinate
- New stack: novel architecture, parallel sampler, and a training method called RLCD (Reinforcement Learning for Calibrated Decisions)
- LLM-level intelligence on System One tasks while being two orders of magnitude faster and more efficient
- Third-party test: 139/140 labels correct, 4.6x faster and 83% cheaper than Luna — though it once gave max confidence to an answer missing currency info
Community demos already show 200ms browser-agent clicks, 16.5 legal docs/second QA, Tetris play, and 1,024-persona crowd simulation.
More from coding & agent
- Diorama gives OpenAI Codex coding agents a visual office you can watch work in real time — davidfromkansas · 2026-09-17
- Code-first, UI on top: building bespoke brand design tools with AI — floguo · 2026-09-17
- Study of 7 models across Claude Code, Codex, Pi: harness barely affects success but swings cost — DavideCrapis · 2026-09-17
- AI trading bot built with Jev is down 85%, owner shrugs it off — generativist · 2026-09-17
- Redditor's 3-Day SoL-Pi Test: Memory Objects Save ~12k Tokens Per Tool Run — Garblyx · 2026-09-17
- Reviewing AI code through Steve Jobs' lens: unseen internals deserve beauty too — sergeykarayev · 2026-09-17