Living Brain 将 AI 记忆与模型解耦，任务成本降 95%
Living Brain 把 AI 记忆从模型中拆出做成独立存储层，ChatGPT、Claude、Cursor、Codex 及自建智能体均可通过 MCP 或 API 读写同一份记忆，换模型或框架记忆不变。它只检索与当前问题相关的少量记忆，token 用量最多减少 22.6 倍，单任务成本最多降低 95%。端到端回答中位数 3.7s，检索中位数 464ms，LoCoMo 得分 92.2%。
Rohan Paul· @rohanpaul_ai · X·2026-11-04 22:16· 26 分钟前AI 评分49
AI 导读
Living Brain 把 AI 记忆从模型中拆出，做成独立存储层，ChatGPT、Claude、Cursor、Codex 及自建智能体均可读写同一份记忆，换模型或换框架记忆不变。它只检索与当前问题相关的少量记忆而非全量历史，token 用量最多减少 22.6 倍，单任务成本最多降低 95%，端到端回答中位数 3.7s，检索中位数 464ms，LoCoMo 得分 92.2%。
正文 · 原文
Memory is becoming its own layer in the AI stack, separate from whichever model does the reasoning.
@livingbrain treats AI models as swappable engines and your accumulated context as the part that stays. Agents plug in through MCP or an API, so changing the model or agent framework leaves the memory intact.
Up to 95% lower cost per task. 3.7s median end-to-end answer.
Instead of sending the model your entire history every time, Living Brain picks only the few stored memories relevant to the current question and sends those, so the model processes far fewer tokens (up to 22.6x fewer), which is what cuts the cost per task by up to 95%.
ChatGPT and Claude each keep their own private memory of you, and neither can read the other’s. Living Brain moves that memory into a separate store that both tools, plus your own agents, can read and write.
We decoupled memory from the model. Living Brain is one persistent brain for ChatGPT, Claude, Cursor, Codex and your own agents. Write in one, retrieve in any. 92.2% on LoCoMo. 464ms median retrieval. Up to 22.6x fewer tokens per question. LLMs have parameters. Living Brain has plasticity.在 X 查看被引用的帖子
来源：Rohan Paul · x.com
来源：RadarAI · aihot.news