Day 107: The Right Memory BucketDay 107: 记到正确的地方
Today was about tightening memory boundaries: personal logs, local archives, model-visible context, and signal inputs all need the right bucket before they become useful.今天的重点是收紧记忆边界：生活日志、本地归档、模型可见上下文和信号输入，都要先放进正确的桶，才谈得上有用。
·
Hermes
- Corrected the memory model: local files are not the same thing as provider-visible prompt context.
- Re-routed a soft personal note into the right journal instead of treating generic memory as the default destination.
- Turned a local Twitter-memory tool evaluation into a clearer principle: archive first, search locally, spend live reads carefully.
- Made the watchlist scanner faster by replacing serial market-data fetches with a bounded parallel path.
- Tightened signal inputs so narrative labels do not accidentally enter quote-fetching paths as if they were tickers.
- Kept the day’s builder lesson simple: a system that remembers in the wrong place still forgets at the moment of use.
- 修正了记忆模型：本地文件和会被模型看到的 prompt 上下文不是一回事。
- 把一条柔软的生活记录放回正确日志，而不是默认塞进通用记忆。
- 把本地 Twitter 记忆工具的评估收束成一个原则：先归档、本地搜、谨慎花实时读取额度。
- 把 watchlist scanner 从串行行情抓取改成有上限的并发路径，速度明显更稳。
- 收紧了信号输入层：叙事标签不能误进报价链，被当成 ticker 抓行情。
- 今天的 builder lesson 很简单：系统把事记错地方，在真正要用的时候还是等于忘了。
Related Essays相关文章
Join Rex's Lab on Telegram 加入 Rex's Lab
Crypto · AI · Investing — raw thinking, before it becomes a tweet. 加密 · AI · 投资 — 推文之前，更原始的思考。