Governed retained-state selection: making agent memory actually change decisions, not just recall
nice2Bnice2 · reddit · 2026-10-23
A Reddit deep-dive argues that agents remember mistakes yet repeat them because retrieving information and letting it influence decisions are different things. The proposed CAAI approach, governed retained-state selection, lets retained history directly reweight candidate actions downstream of generation—no RAG required—while explicit rules can veto unacceptable choices.
Claimed benefits (unverified): fewer repeated model calls, better behavioral continuity, inspectable decisions. Ongoing 'Evolution 2' work targets structured continuity—what matters, what changed, when old info should stop influencing decisions—with applications in autonomous agents, simulation and game NPCs. Math paper and reference implementation are public; the commercial middleware is proprietary.
More from coding & agent
- Claude cut a 2.6-hour video export to 20 minutes — but only after being told to rebuild the server — byhewar · 2026-10-23
- 3 flight-booking rules my AI agent checks that I'd never do by hand — Efistoffeles · 2026-10-23
- OpenJet claims Qwen3.8-27B coding runs 1.9x faster than llama.cpp on one RTX 3090 — JeremyCMorgan · 2026-10-23
- Building an autonomous SEO agent in Hermes with three skills: programmatic pages, Keywords Everywhere, DataForSEO — EXM7777 · 2026-10-23
- Building a sports betting AI agent with onchain execution and human approval by default: architecture and lessons — Dry-Comfortable-2678 · 2026-10-23
- How do you audit AI agent decisions in regulated workflows? A field survey — Ok_Way_1993 · 2026-10-23