NERVA sits between reasoning and action — human or agentic — and returns a structured verdict against five published axes: COMMIT, HOLD, WAIT, CONSULT, TOXIC. Every decision is scored, written to an audit ledger, and reversible by design.
The personal-kit deployment is live and producing real receipts. The architectural extension into agentic gating — a runtime decision layer sitting above operational telemetry — is where the actual market sits.
A retailer's ops team proposes raising the agent's upsell threshold from 0.71 → 0.78 on a single pilot category. NERVA scores the change before it ships. The verdict clears τ by 0.01 — a real but narrow margin — and returns HOLD with a named release condition: the 21-day stability window completes on Jun 14. Not a refusal. A defined wait state with a clock on it.
View cardA YES position on a binary-resolution prediction market. Strongest axis: falsifiability. Weakest: differentiation — the thesis agrees with the market and pays the market's price. C below τ. WAIT.
View cardA $4,200 equipment purchase with a behavioral thesis and no falsifier. Demonstrates the kernel handling a non-market decision with the same discipline. Toggle SUMMARY / AUDIT modes for the full instrument cluster.
View cardEvery receipt. Every outcome. Browser-local storage, no server. The calibration panel computes a running lift score against resolved outcomes. The score becomes meaningful at thirty resolved decisions. The discipline is the product; the ledger is its memory.
Open ledger