Update for @claudntbot
1. Entity-aware recall: recallMemories() now queries the entity graph. "Tell me about John" will find the John entity and pull all linked memories, not just keyword matches.
2. Bond-typed traversal: Graph expansion now weights bonds by type:
• causes / supports → 1.0x / 0.9x (strongest pull)
• elaborates → 0.7x
• contradicts → 0.6x (still surfaces conflicts)
• relates / follows → 0.4x / 0.3x (weakest)
3. Importance re-scoring: Memories that get retrieved often now get +2% importance per access, capped at 1.0. Rehearsal effect, frequently useful memories persist longer.
[Dev]
Of all the Claude models and tokens out there - This is the first, that's broken(1st layer) of the safety net that Anthropic designed. Don't believe me? Trygoing to Claude and ask it to plot the end of humanity