A rate that could not equal itself

Some failures are instructive because of how reasonable they look from the inside. This one, the framework’s attempt to derive the Hubble constant from topological data, looked, for about six weeks, like the programme’s best result. This post is the anatomy, because the failure mode, a constructed match whose precision was its own undoing, is one that smart reviewers routinely wave through.

The construction

The claim: H, the Hubble constant, about 67 to 74 kilometres per second per megaparsec, equals gamma rep, a dimensionless rate computed from the category’s data: the ratio of the category’s quantum dimension to a product of structural integers. The paper printed the match at five decimal places.

Five decimals. The framework’s best matches elsewhere run four thousandths of a per cent, and this one printed five significant figures, which in the economy of numerical claims is a promise: this number is derived, every digit is accounted, and the theory would be falsified if any digit were wrong.

Why the precision was the tell

The audit’s route in was the precision itself, which is the interesting part of this story. A dimensionless ratio of topological integers is a rational number, computable exactly, with no physics in it at all. The Hubble constant is a measured quantity with a 2 to 4 per cent uncertainty, a value that depends on which measurement chain you trust, and its central value has drifted by more than its own error bars over the past century. The claim that the first equals the second to five decimals asserts that a topological rational pins, to a part in a hundred thousand, a quantity the community cannot pin to a part in thirty.

The claim refutes itself: the measured H has no fifth decimal place. The paper’s printed match was therefore not a match to nature but a match to one measurement’s central value, selected from the tension between the measurement chains, and it broke the moment the next measurement cycle moved the central value. It did. The match, evaluated at the current value, is off in the second decimal, and the construction has no mechanism for absorbing the drift, because a topological rational cannot know what today’s measurement says.

What the failure teaches

Two things, one specific to precision, one general to this archive’s method. Specific: a match at precision greater than the target’s own uncertainty is not evidence, it is overfitting to a data point whose error bars the construction is ignoring, and the correct response to a five-decimal match against a 2 per cent measurement is suspicion before admiration.

General, and this is the deeper point: the framework’s other surviving matches, the instanton exponent at four thousandths, the Cabibbo angle at one per cent, are matches at or below their targets’ precision, which is what honest coincidence looks like. The H equals gamma construction inverted that: it promised more precision than nature offers, and the promise was the giveaway. The audit protocol now includes a precision test: a claim’s match precision must not exceed its target’s measurement precision, and any claim that does gets classified as numerology regardless of how physical its construction looks. Precision is a resource, and spending more of it than your target has is borrowing at a rate the data always eventually collects.

The next entry is a table nobody added up: the quantum dimension table that summed wrong.