Jev 背景梳理:OpenAI 前员工做的 System One 决策模型,4000 万美元种子
Diogo Almeida 离开 OpenAI 后 stealth 两年做出 Jev,不做文本生成,只把非结构化状态转成带校准概率的类型化决策。
SITUATION EXPLAINED: "Models have been superhuman at chat for years, so where is all the automation?" • Diogo Almeida spent years at OpenAI on the instruction-following work behind ChatGPT, then two years in stealth building an answer • His model, Jev, takes unstructured state and returns typed decisions with calibrated probabilities, in one parallel pass rather than token by token • It cannot generate text at all • They call it a System One Model, after Kahneman's fast intuitive thinking • It runs 20 to 200x faster and 40 to 400x cheaper than an LLM, at $0.042 per million input tokens with output free • $40 million seed led by DCVC. No independent evaluations at launch @sophiadew: "Maybe we don't need another general purpose model to just talk and talk and talk. We just need something that's really good at not hallucinating and making better decisions."
这条是引用着另一条写的:「ChatGPT 联合发明人发布新模型 Jev」