无需大模型，从 Git diff 中自动推断 AI 写代码的意图，还能找回消失的聊天记录。
Article URL: https://github.com/HUKAIR/CodeTalk Comments URL: https://news.ycombinator.com/item?id=48830345 Points: 1 # Comments: 0
无需LLM调用的Agent记忆层，利用Hebbian规则与模糊偏好图实现低成本、类脑的持久记忆更新。
Hi HN, FERNme is a memory layer for AI agents. Most systems (Mem0, etc.) call an LLM on every turn to extract/summarize memory, or dump everything int…
将任意文档快速转为MCP端点，且查询时无需调用LLM，大幅降低成本与延迟。
Article URL: https://github.com/hashwnath/KMCP Comments URL: https://news.ycombinator.com/item?id=48469712 Points: 2 # Comments: 0
从PRD自动生成可启动的FastAPI应用，无需调用大模型，600行Python实现，开发效率飙升。
Article URL: https://github.com/Anioko/spec-driven-development Comments URL: https://news.ycombinator.com/item?id=48448063 Points: 1 # Comments: 0