Falsification register

A proposal that cannot be killed is not science. Each restoration lever carries a named, measurable falsifier -- SP1 (restore-gain), SP2 (reduce-forcing), SP3 (remove-sensitiser) -- plus a FRAMEWORK-level falsifier on the γ read itself. If the stated experiment comes out the stated way, the corresponding claim is wrong.

VP-SPEC honesty: the register makes each hypothesis refutable. The gate is that every proposal id plus FRAMEWORK has at least one falsifier (True). Naming the kill-condition is what separates a hypothesis from a claim.

S1 -- restore the gain

If restoring feedback gain (re-sensitisation) in a loop that has ALREADY crossed to the disease basin fails to return the state, while only reducing the chronic forcing does, the S1 'deepen-the-basin-alone restores' premise is wrong (gain restoration is insufficient once crossed).

S2 -- reduce the forcing

If lowering the chronic forcing abolishes the defended setpoint entirely (the loop holds NO setpoint without the forcing -- no bounded return to a healthy value), the S2 'controlled return to a healthy setpoint' premise fails.

S3 -- remove the sensitiser

If removing the upstream uncoupling/inflammatory program does NOT raise the crossing threshold (spinodal) of the disease basin in an independent loop-gain assay, the S3 'remove-the-sensitiser' premise is wrong.

FRAMEWORK -- the read itself

If the gamma-derived |h_sp| ordering of the setpoint-node set is uncorrelated with ANY independent promoter-switch readout, the read's organising relevance to these loci is weakened (the read remains [V] as a number, but its claim to order the nodes would not hold).

Why this is the last chapter

The map ends on its own kill-conditions deliberately. The foundational separation, the torpor switch, the disease law, and the restoration levers are all offered as REFUTABLE structure -- the honest end of a research program, not a conclusion.