● 16-Brain Fusion — Genomics, Transcriptomics, Pathway, GCN, Chemistry, Clinical, Biomarker, DepMap, PubMed, Cancer, Competitor, Curiosity, Quantum, Evidence, NPEAD, Proteomics
● Online Learning — Learns from every outcome (corrections applied to weights each run)
● GCN Self-Teaching — Graph neural network trains on CIFOG nodes → writes signals back to brains
● Teacher (Claude Opus 4.8) — Evaluates brain quality, generates weight corrections nightly
● Repurpose-First Reasoning — Checks existing drugs before de novo generation
● Quantum Meta-Learning — VQE search proposals for ambiguous scoring cases
● Real Omics Refresh — 441 signals from TCGA (4 cancers) refreshed each run
● Autonomous Discovery — Promotes high-scoring genes to heroes without human input
● Evidence Hash Chain — Every decision SHA-256 linked, S3 immutable ledger
● Authority Gating (PQM) — Self-checks permission before acting
● S3 Weight Backup + Snapshots — Full state reproducibility per run
● Gene Discovery Scan — Queues novel candidates from platform state
◐ Active Learning (SMARTER_COMPOUNDS) — MD survivors bias next-gen molecules
◐ Preliminary Quality Gate — Druggability scoring before GPU spend
● Brain Disagreement Surfacing — 5-brain quorum voting, conflict detection, severity scoring
● NPEAD v2 (6-Axis) — TV, SV, UR, F, S, EC + Default/Hypothesis/Exploration modes
● BioBrain Phase 1 — SafetyNet + SynthNet + BindNet + PocketNet + Orchestrator
● BrainSignal Reasoning — Structured "why" field per gene (Level 4 foundation)
● Discovery Outcome Recording — Failure reasons + scaffold stats + generator version
◐ Decision Comparison (v1/v2) — A/B framework ready, awaiting hypothesis engine
○ Hypothesis Generation — Proposes "test X because Y" (Level 4, 3 days)
○ Biological World Model — Causal graph of mechanisms (Level 5, 7 days)