用 Astra 智能体数量推演 AGI 后的递归自我改进

Haider. @haider1原推发布于 2026.09.10

在 X 上打开原推

if you still can't see what AGI would mean at scale lemme put it simply: openai has already reported 3.1 agent workdays per human researcher workday now imagine that with AGI: 2,000 human researchers + 20,000 Astra-level agents doing AI research nonstop, and 20,000 is easily conservative -- with enough compute, that could scale to 200,000 agents that's when RSI starts