What a no-go theorem actually buys you

A no-go theorem proves that something cannot be done, and the ordinary reaction to one is disappointment. This post is the case that in a research programme, no-gos are among the highest-value results you can produce, and that three of the strongest outputs here are doors that closed.

The three doors

The first: the vacuum braiding no-go. The framework wanted the Standard Model’s mixing matrices to come from the category’s braiding data, the way charge quantisation comes from topology. The theorem: the braid generators for this category are rigid algebraic numbers, and the mixing they produce is wildly wrong, an off-diagonal element around 0.46 against an observed 0.225, with no freedom to tune. What the no-go bought: the end of a dead end and the redirect that became the modular-flavour paper, which gets its mixing from holonomies on a torus instead, where the numbers land within one per cent.

The second: the Yukawa no-go. The same category, asked to explain mass hierarchies through fusion multiplicities, gives every admissible Yukawa triple multiplicity one. The hierarchy cannot come from the category alone; the theorem says so exactly. What it bought: the recognition that hierarchy must come from somewhere else, the modular weights and the background modulus, and a statement of where to look that is precise enough to fail.

The third: the third-order interference exclusion, from the walkthroughs. No nonlinear multiverse, because nonlinearity is visible in a three-slit experiment that has already bounded it to one part in a hundred. What it bought: a whole class of tempting models, branch couplings with nonlinear structure, ruled out before anyone built one.

Why closed doors are worth more than open ones

An open door costs you time in proportion to how attractive it looks. The braiding route to mixing looked excellent: topology to mixing with no parameters, exactly the kind of derivation the framework exists to find. Every week spent on it after the multiplicity table was computed was a week spent deriving a wrong number more carefully. The no-go converted that cost into a boundary: not a suspicion that the route fails, a proof, with the specific rigidity that fails it. Suspicion wastes attention forever. Proof ends the conversation and moves the labour to the next route.

There is also a subtler economy. A no-go constrains the theory’s escape routes, and theories are mostly made of escape routes. The Yukawa no-go says the category alone cannot do it, which means any successful account must import something, and imports are visible and criticisable in a way that quiet modelling choices are not. The theory got smaller and more falsifiable in the same stroke.

How to earn a no-go

The recipe, from doing it badly first: state the mechanism as precisely as the claim you hoped to prove, find the quantity that would have to vary for the mechanism to work, and check whether the structure pins it. Here the quantity was a mixing entry; the structure was the braid representation; the pinning was exact. If the quantity is not pinned, you do not have a no-go, you have an inconvenience, and the honest output is a benchmark, not a theorem. The distinction matters because a no-go is the only result in the programme that cannot be audited into weakness. It is already the weakest, most robust thing on the shelf: a door that is shut for reasons that do not depend on any measurement.

The next post takes the same discipline into theory, where the rules get written before the result exists: preregistration for theory.