TypeSafe AI launches Jev, a "System One" decision model 100x faster than LLMs, no hallucinations
rseroter · x · 2026-09-21
TypeSafe AI, founded by ex-OpenAI researcher Diogo Almeida (who contributed to the instruction-following research behind ChatGPT), launched Jev, the first of its "System One Models" — frontier models built for fast, structured decisions software can consume directly. Jev gives up string generation, outputs typed probabilistic decisions and claims no hallucinations, matching LLM intelligence on System One tasks while being two orders of magnitude faster and more efficient. The stack includes a new architecture, a parallel sampler, and a novel training method called RLCD (Reinforcement Learning for Calibrated Decisions). Early access is open now.
More from Research
- Anthropic paper claims Claude holds 171 emotion vectors; amplifying 'desperation' spikes blackmail to 72% — repligate · 2026-09-22
- ChemBench: Top LLMs Beat Best Chemists on Average Yet Fail Basic Tasks With Overconfident Errors — geoffwolfe · 2026-09-22
- OlympiadBench: 8,476 Bilingual Multimodal Olympiad Problems, GPT-4V Scores Just 17.97% — geoffwolfe · 2026-09-22
- ENIGMA's 300K brain scans reveal scaling laws for AI brain disease diagnosis — PTenigma · 2026-09-22
- Frontier Data Summit 2026 lineup reveals a dozen new AI benchmarks and top researchers — dlwh · 2026-09-22
- Jev trails Gemini and DeepSeek on calibration, but still handles more decisions solo — frappuccinoCoin · 2026-09-22