The Titan Neural Cortex (tools/titan/neural_cortex.py, ~600 lines) is a set of 6 micro-networks (~77K parameters total) that make Titan's memory storage intelligent.
6 Modules
| Module | Params | Function |
|---|---|---|
| MIS | ~12K | Memory Importance Scoring — how important is this memory? |
| ET | ~10K | Emotion Tagging — what emotional valence does this memory carry? |
| RWL | ~15K | Retrieval Weight Learning — how should retrieval signals be weighted? |
| AS | ~13K | Associative Strengthening — Hebbian "fire together, wire together" |
| CG | ~14K | Consolidation Gating — should this memory be consolidated during dreaming? |
| ID | ~13K | Interference Detection — does this new memory conflict with existing ones? |
All modules run on CPU-only PyTorch. Training happens during Dream Daemon consolidation phases. The cortex is the reason Titan's memory grows smarter over time — it's not just a database, it's a learning memory system.