The VIHAAN protocol: discipline over chaos
The authority layer. When uncertainty rises, autonomy is deliberately narrowed — and the system has to earn it back in stages.
VIHAAN is the final authority layer, and the design principle it encodes is the one I would keep if I had to throw away everything else in the system:
When uncertainty rises, autonomy should narrow — not widen.
The instinct runs the other way. A system in trouble looks like a system that needs more capability brought to bear. In practice, trouble is exactly when your models are least reliable, because trouble means conditions have left the distribution they were trained on. The moment you most want intelligence is the moment you can least trust it.
Escalation
normal → warning → critical → vihaan_activated
Deterministic, driven by sustained systemic risk signals rather than instantaneous ones. A spike does not trigger VIHAAN; a spike that does not resolve does.
What activation does
On activation VIHAAN applies authoritative overrides:
- Disables MORL and PULSAR
- Caps oracle weight
- Freezes the meta layer
- Forces the baseline decision path
The system is stripped back to the deterministic controller that existed before any learning was added. This is why that controller was built first, and why it is maintained as a first-class component rather than left as scaffolding — it is the floor the whole structure falls back to.
Earning it back
stabilize → monitor → gradual_reintroduction → full_restore
Recovery is a state machine, not a flag. Intelligence is reintroduced in stages, each gated on evidence of stability rather than elapsed time. The system earns back its own autonomy.
The failure mode this prevents is oscillation: restore everything the moment the metrics look acceptable, destabilise again, activate again. Staged reintroduction with evidence gates is slower and it converges.
Memory
VIHAAN persists memory signatures and audit logs for replay and governance learning. Every decision is traceable after the fact.
That matters most for the activations that turned out to be unnecessary. A protocol that narrows autonomy will sometimes narrow it wrongly, and without a replayable record there is no way to distinguish "VIHAAN saved us" from "VIHAAN panicked" — which would make the threshold impossible to tune honestly.
On the name
It is named after me, which is either the most or least defensible naming decision in the project. The reasoning: this is the layer that encodes what I am willing to let the system do without asking. If any component should carry the name of the person responsible for it, it is that one.