Jev 创始人访谈：决策模型如何嵌入自主 agent
前 OpenAI 研究员、Jev 创始人谈为何当前 LLM 仍围绕人类设计、RLHF 优化了什么，以及 Jev 如何以更快更便宜的方式嵌入自主 agent。
Jev Founder (ex-OpenAI): x200 faster, x400 cheaper and no human constantly sitting in the loop Diogo Almeida believes models like JEV are what comes after the current generation of AI assistants · 04:10 why current LLMs are still built around humans · 10:35 what RLHF actually optimized models for · 17:50 why software needs decisions, not more text · 25:20 how JEV fits inside autonomous agents · 32:10 what the next generation of LLMs could look like This is not really another “better chatbot” story. The idea is to move intelligence deeper inside software, where thousands of small decisions can happen without constantly returning to a human. One of the most interesting JEV talks to watch right now. Watch the 36 minutes, then read the full Jev Engineer breakdown below ↓
这条是引用着另一条写的：x.com/0xmorlex/status/2101268094387704311