Slow mode
A budget engine operating mode that applies full retrieval depth, all agent roles, cross-encoder reranking, and maximum reasoning complexity. Used when the utility of a decision justifies the computational cost. The default for high-stakes or high-uncertainty operations.
Slow mode is the full-capability operating mode of the cognitive loop. All five agent roles are dispatched, retrieval uses maximum depth with hybrid BM25 plus k-NN search and cross-encoder reranking, the world model runs full prediction with causal model consultation, and the reflection loop is available to review the outcome afterward. The budget engine selects slow mode when the operation's utility score exceeds the threshold that justifies full processing cost. High-stakes decisions (ones with large predicted reward or high risk), high-uncertainty situations (novel task classes, low world-model confidence), and explicitly flagged critical operations all qualify for slow mode. The contrast with fast mode reveals the budget engine's role: it is not about quality vs. speed in general, but about allocating full-quality processing to decisions where it matters most.