My AI agent kept quoting a number I had already corrected
In August my coding agent kept telling me that one kind of article on this site draws about four times as many comments as another. A week earlier I had corrected that number myself: with more data it was closer to two to one.
The old note was still in its memory. It sounded exactly as sure as the new one, and it kept winning, because nothing about it looked old.
That's the failure I find hardest to catch. A missing memory is loud: the agent asks, you answer, done. A stale one is quiet. It hands you a confident answer that stopped being true, and it reads like every correct answer it has ever given you.
I tried to fix it by ranking older notes lower, and that went badly.
So I'm collecting real cases for the next part of this series:
- What did your agent remember that had stopped being true? A config value, a renamed API, a decision your team reversed, a workaround for a bug that's long fixed.
- How did you notice? A failing build, a reviewer, a customer, luck?
- How long was it wrong before anyone saw it?
Short answers are welcome. The ones with a concrete example are the ones I'll learn the most from.
cachly.dev
For solo developers, teams and whole companies
Explain it once. Every AI knows it.
cachly keeps what you explain to your AI in every session. Claude Code, Cursor, Copilot and Windsurf read your rules, decisions and solved bugs: for you alone, for your whole team and for every team in your company. Servers in Germany, with SSO and a DPA for companies.
Servers in the EU · one command, restart your editor, done · 1 project free
Four ways to install
- Claude Code
/plugin marketplace add cachly-dev/cachly-mcp /plugin install cachly-brain@cachlyThen turn on auto-update once:
/plugin→ Marketplaces → cachly → Enable auto-update- Any MCP editor
npx @cachly-dev/mcp-server@latest autopilot- JetBrains
- JetBrains Marketplace: cachly-brain