MIT 张艾伦谈递归语言模型:Claude Code、Codex 与 Pi 本质相同,以及 OpenAI 万级智能体实验
RT by @swyx: Recursive Language Models: Claude Code, Agent Swarms, Big Research Bets, & the $40M AI Problem Solver https://www.latent.space/p/rlm MIT’s @a1zhang explains why Claude Code, Codex, and Pi are basically the same, how RLMs use code, context offloading, and recursive subagents to generalize across tasks, why one expert can sometimes replace a trillion-token brute-force search, what OpenAI’s 10,000-agent, 130B-output-token experiment reveals about the future of language models, and why academia’s biggest advantage is the freedom to take weird, ambitious research bets.
MIT 研究者 @a1zhang 在 Latent Space 播客中提出递归语言模型(RLM),认为 Claude Code、Codex 和 Pi 本质上是同一类东西,通过代码、上下文卸载和递归子智能体实现跨任务泛化。他援引 OpenAI 一项 10,000 个智能体、130B 输出 token 的实验,讨论语言模型的未来,并指出有时一个专家就能替代万亿 token 的暴力搜索。
来源:X:swyx (@swyx) · x.lingyaoai.com