I've been working on MeshCtx, an open-source AI agent with 17 brain regions and a memory engine that uses FSRS spaced repetition + schema consolidation (episodic→semantic→core) + sleep-phase offline processing.
The key innovation is the ARCHIVAL system that actively prunes memories without deleting them - they can be recovered later.
Benchmark results: - LongMemEval EM 54.2% / judge 83.3% - 16KB budget +8.3pp improvement - Tool compression -95.5%
I'm curious about others' experiences with: 1. Memory consolidation approaches (episodic→semantic→core) 2. Spaced repetition for AI agents 3. Active forgetting/pruning strategies
GitHub: https://github.com/LucyAndLuna2023/meshctx
What patterns have worked for you?