笑死我了 (引用推文要点:递归语言模型——Claude Code、智能体集群、大型研究赌注,以及 4000 万美元的 AI 问题求解器。MIT 的 @a1zhang 解释了为什么 Claude Code、Codex 和 Pi 基本是同一回事,RLM 如何利用代码、上下文卸载和递归子智能体来泛化到不同任务,为什么一个专家有时能替代万亿 token 的暴力搜索,OpenAI 的 1 万智能体、1300 亿输出 token 实验揭示了语言模型的什么未来,以及为什么学术界最大的优势是自由去做古怪而雄心勃勃的研究赌注。)
lmaoooo
【引用 @latentspacepod】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.
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.在 X 查看被引用的帖子
来源:swyx · x.com