2 hours15 research jobs, 10 deltas
Fresh brains study themselves before acting like long-lived companions.
Maturity and status
Progress is evidence-gated.
JARVIS is architecturally complete, runtime maturing. A fresh brain is supposed to start cautious, locked, and low-evidence.
30 min boot cooldown100 shadow decisions at 55% win rate
Policy must beat deterministic behavior before authority expands.
4 hours shadow50 validated predictions at 65% accuracy
Predictions mature from shadow to advisory/active only with evidence.
48 hours shadow100 validated simulations at 70% accuracy
Hypotheticals stay read-only and non-authoritative until proven.
30 min warmup10 positive deltas at 40% win rate
Research and safe-apply authority are earned through outcomes.
7 days stability2+ promoted specialists with >0.05 mean impact
Policy feature expansion waits for long-run specialist evidence.
Common false positives
Do not confuse caution with absence.
Policy NN at 0 decisions after a reset is expected.
World Model Level 0 is normal before enough validated predictions and runtime.
Autonomy Level 0 means the warmup and outcome gates have not been earned yet.
Dashboard red bars often mean pre-mature evidence, not broken architecture.
Default-off rollout flags are governance, not missing features.
Synthetic lanes accelerate perception only without contaminating lived memory.