Capability search
The open-ended evolution process that searches for structural improvements to the cognitive system when policy learning has converged and performance remains insufficient. Triggered by repeated failure, curiosity pressure, or persistent contradictions. Guided by the curiosity engine and constrained by constitutional checks.
Capability search is the mechanism that distinguishes open-ended intelligence from ordinary policy adaptation. Policy learning adjusts weights within the existing architecture; capability search looks for new structural capabilities the architecture does not currently have. It activates only after the developmental engine confirms the system has reached the open-ended phase (mean capability above 0.85) and only when policy learning has visibly converged without resolving a persistent failure class. The curiosity engine's information gain formula prioritizes search directions where uncertainty is high and potential value is large. The dream engine simulates proposed capabilities before they face constitutional review, cheaply filtering out proposals that the world model predicts will make things worse. Accepted capabilities enter the system at low weight and must earn their place through reinforcement.