Agent Saved 80,000 Characters of Lessons — the Next Agent Never Read Them
Mr_ZapatoBlanco · reddit · 2026-09-28
A Reddit user describes how their team auto-saved reviewer feedback for coding agents, accumulating 20 entries (80k characters) in 12 hours — only to find the next agent never read that field, only the coding standards and pinned notes. They had built the writing side of agent memory but no path for lessons to influence the next run. Fix: disable auto-append by default, make rule extraction and task linking separate steps, and ask where/when info is read and how to verify it prevents repeat mistakes. The real work, they conclude, isn't saving conversations but deciding what should affect the next decision.
More from coding & agent
- Google Antigravity adds native inter-agent messaging and user-to-agent DMs — rseroter · 2026-09-28
- Dataviz expert says AI pushes him toward lower-level charting languages for finer control — randal_olson · 2026-09-28
- Cloudflare Birthday Week 2026: native Rust in Workers, agentic browser, EmDash CMS — ritakozlov · 2026-09-28
- Creative arbitrage: build a Claude Code skill to ship 500+ static ads for almost nothing — alexgoughcooper · 2026-09-28
- Git is the cheapest insurance for your Obsidian vault, says automation guide author — dSebastien · 2026-09-28
- Cloudflare open sources Forge: one pipeline to generate SDKs, CLIs and docs — dee_hw · 2026-09-28