Remnic blog
Notes on agent memory, correction, and trust from the Remnic project. Subscribe with the RSS feed.
Dreams: how Remnic consolidates memory in its sleep
Remnic's background consolidation runs in three named phases: light sleep, REM, and deep sleep. Here is what each one actually does.
July 18, 2026Engram to Remnic, three months later
A rename retrospective: what moved cleanly, what still answers to the old name, and how we are closing the gap without breaking anyone.
July 18, 2026MemCorrect: benchmarking what happens after you say 'that's wrong'
Recall benchmarks measure what a memory system finds. MemCorrect measures whether a correction sticks, and what it breaks on the way.
July 18, 2026Procedural memory is now on by default
Remnic now captures how you do things, not just what is true, and injects the runbook when a matching task starts. No configuration required.
July 18, 2026Remnic Relay: correct once, every agent learns
What we built for OpenAI Build Week 2026: memory that heals an agent team instead of quietly poisoning it.