Root cause of the token-cost/scattergun symptom. Hypatia.LearningScheduler (supervised, 5-min poll) ingests outcomes but never calls Hypatia.Neural.Coordinator.force_cycle/0; Coordinator.init/1 never calls Hypatia.Neural.Persistence.load_all/0 (zero callers). Result: ESN/RBF/MoE run on init-defaults forever; pattern_analyzer defensively ignores untrained output. Training data is ALREADY sufficient (outcomes: recipe-shell-quote-vars 166-pt series ≥60; patterns/registry.json 9 ≥5) — only the trigger is missing. Fix: LearningScheduler→force_cycle after ingest; Coordinator.init→load_all+hydrate (RBF/MoE/trust restore; ESN/LSM persistence is summary-only — lossless ESN persistence tracked as follow-up). Parent #273.
Root cause of the token-cost/scattergun symptom.
Hypatia.LearningScheduler(supervised, 5-min poll) ingests outcomes but never callsHypatia.Neural.Coordinator.force_cycle/0;Coordinator.init/1never callsHypatia.Neural.Persistence.load_all/0(zero callers). Result: ESN/RBF/MoE run on init-defaults forever;pattern_analyzerdefensively ignores untrained output. Training data is ALREADY sufficient (outcomes: recipe-shell-quote-vars 166-pt series ≥60; patterns/registry.json 9 ≥5) — only the trigger is missing. Fix: LearningScheduler→force_cycle after ingest; Coordinator.init→load_all+hydrate (RBF/MoE/trust restore; ESN/LSM persistence is summary-only — lossless ESN persistence tracked as follow-up). Parent #273.