MetroVolt's method is gate adjudication: each governing uncertainty is assigned to a named gate with pre-registered acceptance criteria, and design credit is only booked on the side of the gate that has already closed.
Four decision gates anchor the register — confinement (S18), the hot-ion two-fluid balance (S26), and the DEC exhaust chain (S51/S65) — plus seven pre-registered predictions (P1–P7) that the G1 testbed will adjudicate. Where a gate has closed against ambition, the design moved: S26/S72 demoted hot-ion operation to gated upside and made near-thermal the baseline.
The mechanism's teeth are pre-registration: criteria are frozen and deposited before results exist, so neither we nor our reviewers can move goalposts afterward. The register's DS-tier rows (design-space) are barred by rule from altering the frozen point they explore.
Investors call this staged risk retirement; scientists call it falsifiability; both fund it more readily than confidence. The gate ledger also makes progress legible — each closure is a discrete, verifiable value step, not another optimistic press cycle.
| Named gates | S18 · S26 · S51/S65 (+ P1–P7 at G1) |
| Criteria | pre-registered, deposited before results |
| Design consequence example | hot-ion → gated upside (S26/S72) |
| Design-space rule | explore only; frozen point immutable |
| Where value moves | one gate closure at a time |