Why We Freeze a Design Point

Freezing turns a bag of loose figures into one coherent machine, so analysis, documents, and predictions all refer to the same thing.

Coherence is the whole point

A design point is only meaningful if its numbers are mutually consistent. Fusion parameters are tightly coupled, so a figure updated in isolation can contradict five others without anyone noticing. Freezing forces the whole set to be reconciled at once, then held, so that every downstream use starts from a coherent machine rather than a snapshot of a work in progress.

What breaks without a freeze

• Documents drift: two decks quote two slightly different currents and both feel authoritative.
• Analyses rot: a study built on a superseded field is invalid but still circulating.
• Predictions blur: you cannot pre-register a prediction against a target that moves.
• Comparisons fail: a later measurement cannot be checked against an undefined baseline.

Freezing enables falsifiability

A pre-registered, falsifiable prediction requires a fixed prediction. You cannot honestly say a measurement would refute the design if the design can be edited after the fact. The freeze is what makes the prediction binding: the numbers were set, dated, and published before the test that could disprove them.

This is also why we resist the urge to keep polishing. Continuous quiet improvement feels virtuous but destroys accountability. A frozen, dated, imperfect design that anyone can check against reality is more honest than a perpetually updated one that can never be pinned down or refuted. The discipline, in short, is to choose a coherent and testable design over a perpetually improving but unaccountable one, because only the first kind can ever be honestly checked against the world.