tl;dr: TypeSafe AI, founded by ex-OpenAI researcher Diogo Almeida, has launched Jev, a "System One Model" designed for fast, structured decision-making rather than text generation. Trained via Reinforcement Learning for Calibrated Decisions (RLCD), Jev produces type-safe structured outputs with calibrated confidence scores, claiming to be 40-200x faster and up to 444x cheaper than frontier LLMs on comparable tasks, with no hallucinations or type errors. It's aimed at automation workflows, classification, routing, and real-time applications where LLMs are too slow or unreliable to embed in code.