Do you do anything to manage the size of this memory or is that left to the user? I think it would be interesting to make it behave more like real memory and have a way to degrade gracefully, e.g., consolidate old memories to key words only?
Using something similar in my OpenClaw setup — hierarchical memory with L1/L2/L3 layers plus a "fiscal" agent that audits before writes. The gate is definitely the right abstraction.
One thing I've noticed: the "distillation" decision (what deserves L2 vs stays in L1) benefits from a second opinion. I have my fiscal check summaries against a simple rubric: "Does this change future behavior?" If yes, promote. If no, compact and discard.
Curious if you've experimented with read-time context injection vs just the write gate? I've found pulling relevant L2 into prompt context at session start helps continuity, but there's a tradeoff on token burn.
Great work — this feels like table stakes infrastructure for 2026 agents.
Do you do anything to manage the size of this memory or is that left to the user? I think it would be interesting to make it behave more like real memory and have a way to degrade gracefully, e.g., consolidate old memories to key words only?
Random Tuesday thoughts. Neat work!
thanks! lmk if you have any feedback
Using something similar in my OpenClaw setup — hierarchical memory with L1/L2/L3 layers plus a "fiscal" agent that audits before writes. The gate is definitely the right abstraction.
One thing I've noticed: the "distillation" decision (what deserves L2 vs stays in L1) benefits from a second opinion. I have my fiscal check summaries against a simple rubric: "Does this change future behavior?" If yes, promote. If no, compact and discard.
Curious if you've experimented with read-time context injection vs just the write gate? I've found pulling relevant L2 into prompt context at session start helps continuity, but there's a tradeoff on token burn.
Great work — this feels like table stakes infrastructure for 2026 agents.