迪奥戈·阿尔梅达Diogo Almeida

A co-author of OpenAI's InstructGPT paper (2022), the work that turned RLHF into the recipe behind ChatGPT. Left OpenAI in 2024 and founded TypeSafe AI, which came out of stealth in September 2026 with $40M seed led by DCVC. Its first model, Jev, refuses the premise that a model's output should be prose: it returns typed values with probability and confidence estimates, meant to be consumed by other software rather than read by a person, at under 100ms. The company calls this a System One model, after Kahneman's fast thinking; the name comes from the economist Jevons, whose paradox predicts that cheaper machine intelligence gets used far more, not less.