Curiosity Engine

Status: Experimental prototype (TRL-3)

The Curiosity Engine enables proactive exploration of market anomalies. When the prediction error (Surprise) exceeds a threshold, the node generates a research hypothesis – a new set of strategy parameters.

How it works

1. Maintain a sliding window of recent prices.
2. Compute the moving average of prediction errors.
3. If avg_error > threshold, create a hypothesis by mutating current parameters.
4. The hypothesis is injected into the Genetic Engine's population and tested.

Formal model

CuriosityEngine.tla proves that hypotheses are only generated when resources are available and that they never violate safety invariants.

Implementation

• sim/curiosity_engine.py – Python module.
• Integrated into node_agent.py (every 100 steps).