Technical White Paper
How BlueGecko Remembers: A Biological Approach to AI Memory
Technical white paper · March 2026
Current AI assistants treat memory as an append-only log — a flat list of facts that grows without limit and has no notion of relevance, decay, or consolidation. BlueGecko takes a different approach, modeling memory after the biological processes that govern human recall.
Every memory entry carries a vitality score computed via an Ebbinghaus-inspired forgetting curve: vitality decays exponentially with time but is boosted by access frequency and inherent importance. A nightly consolidation process — running at 2 AM while the user sleeps — applies decay, promotes high-value memories, merges duplicates, and prunes entries that have faded below a viability threshold.
The result is an AI that doesn't just remember — it remembers the way you do. Important things stay. Trivial things fade. The relationship deepens over time, naturally.
Contents
- 01The Problem with Flat Memory
- 02The Ebbinghaus Forgetting Curve — Applied to AI
- 03Memory Tiers: Core, Active, Consolidated, Faded
- 04The Vitality Formula
- 05Nightly Consolidation: How It Works
- 06Access Tracking and the Recall Boost
- 07Privacy by Design: On-Device Only
- 08Results and Observations
- 09Future Work: Company-Wide Memory (The Brain)
Get notified when the paper drops.
We'll send it straight to your inbox.