TypeSafe AI Launches Jev: Decision Model That Cannot Hallucinate

TypeSafe AI launched Jev, a decision-making model trained with RLCD instead of RLHF, avoiding hallucination by not generating free text. Jev outputs structured decisions with epistemically honest probabilities. It targets classification, routing, and scoring tasks, claiming equal intelligence to frontier LLMs on System One tasks at a fraction of cost and latency. Output tokens are free.