Proto-Cognitive Architecture v4
Neural Field + Hebbian Memory + Semantic Retrieval + Cognitive Router
Teach the agent facts, then ask questions. The agent uses a continuous Hebbian attractor field to detect cognitive resonance (familiarity) with learned content, and routes queries accordingly:
| Route | Meaning |
|---|---|
| 🟢 CONFIDENT | Strong field resonance + reliable facts → full answer |
| 🟡 CAUTIOUS | Field recognises domain but weak retrieval → hedged answer |
| 🟠 UNCERTAIN | Some retrieval but low familiarity → answer with caveats |
| 🔴 DEFER | No stored knowledge → admits ignorance |
Key properties: Zero forgetting · Paraphrase robustness · O(1) memory · Contradiction handling · Introspectable routing decisions
One fact per line. Facts are encoded into the Hebbian field and pinned to the semantic store. Teaching a contradictory fact automatically supersedes the old one.
Field State
Field norm: 0.000 | Active regions: 0 / 0 protected Attractors: 0 | Steps: 0
Episodic Store
Total: 0 | Active: 0 | Superseded: 0
Cognitive Router
No queries routed yet
Token Usage
Teach: 0 calls (0 tok) | Gen: 0 calls Avg in/gen: 0.0 | Avg out/gen: 0.0