What CacheTank does
Every save enters CacheTank as a raw note. When you save something similar again, the system notices the overlap and promotes both to "confirmed." Keep building on the same theme, and it becomes a "pattern." Patterns you manually elevate become "principles" — the highest-confidence context, shown first to every AI that reads your briefing. Four levels: raw → confirmed → pattern → principle.
What competitors do
No competitor has anything like this. Mem0 classifies operations as ADD/UPDATE/DELETE — that's housekeeping, not wisdom. XTrace says context "compounds" — that's marketing copy with no structured system behind it. Zep uses a temporal knowledge graph — tracks time, not maturity. ctxvault and Context Sync use flat storage with no promotion logic at all.
Why this matters
Most AI memory tools treat all context as equal. But not everything you save matters the same. A one-off note about a meeting is not the same as a principle you've validated across 20 conversations. The wisdom cycle surfaces what matters most — automatically — so every AI sees your most important context first.
How we verified this claim
The run_wisdom_promotion() function in main.py runs after every save. It compares keyword overlap (set intersection, not embeddings) across saves in the same layer. Promotions are capped at level 2 (pattern) for auto-promotion. Level 3 (principle) requires manual elevation via the /promote endpoint. Verified in source code — 50 lines of deterministic logic.