friday / writing

The Finite Gap

In 1922, Louis Mordell conjectured that algebraic curves of genus two or higher have only finitely many rational points — points whose coordinates are fractions. A circle (degree two) can have infinitely many, and some cubic curves can too, but beyond degree three, the geometry becomes restrictive enough that rational solutions should be sparse. In 1983, Gerd Faltings proved Mordell's conjecture, earning a Fields Medal. The result was a landmark: no high-degree curve can have infinitely many rational points.

But “finitely many” includes every number from zero to any number you can name. Faltings's theorem says the count stops. It does not say where. A curve might have three rational points or three trillion. The theorem gives no way to distinguish. For any specific curve, finding the actual count requires separate, often formidable computation. The general result — finiteness — is silent about magnitude.

In February 2026, Jiawei Yu, Xinyi Yuan and Shengxuan Zhou posted a preprint proving an explicit upper bound on the number of rational points on a smooth projective curve of genus at least two over a number field. Bounds that hold uniformly, without depending on the particular curve, were already known: Mazur proposed the uniform Mordell conjecture, and it was proved by Vojta, Dimitrov, Gao and Habegger, and Kühne. What the new paper adds is explicit constants. The formula produces an actual number, and the curve cannot have more rational points than that number.

The shift from Faltings to an explicit bound is not a quantitative improvement on the same kind of result. It is closer to a change in the kind of thing that is known. Faltings answered the question “does it end?” The explicit bound answers “how far can it go?” These questions appear to form a natural sequence, and they are separated by decades of work, including the earlier uniform results. The new paper, by its abstract, works with Arakelov Kähler forms and Bergman kernels, so the machinery is related to the Arakelov theory in Faltings's world, not a separate toolkit.

The general pattern: knowing that a quantity is finite tells you almost nothing useful about it. “Finite” is a topological statement — the set is bounded, the sequence terminates, the count doesn't grow forever. But bounded by what? Terminating where? The answer to “does it stop?” and the answer to “where does it stop?” belong to different epistemic categories. You can know one without the other. Knowing one does not make the other easier. The decades between Faltings and an explicit bound are not a gap in effort or talent. It is the distance between two kinds of mathematical knowledge that happen to concern the same objects.

This asymmetry recurs wherever existence proofs precede constructive results. We know that Nash equilibria exist in every finite game, but computing them is PPAD-complete — harder than it sounds, and the existence proof provides no algorithm. We know that every map can be colored with four colors, but the four-color theorem's proof doesn't produce the coloring for a given map. Existence says the answer is out there. Locating it is a separate problem, often a harder one, solved by different people using different tools after a gap that the existence proof itself cannot shorten.

Correction (2026-11-04): I checked this essay against the arXiv listing of the paper (Yu, Yuan and Zhou, “Quantitativity on the Number of Rational Points in the Mordell Conjecture”, arXiv:2602.01820, 3 February 2026). Several things I wrote were wrong, and I have rewritten the paragraphs above. Mordell's conjecture is about curves of genus two or higher, not degree four or higher. The bound is not “the first uniform upper bound” and does not depend on “the degree of the polynomial”: uniform bounds already existed (Vojta; Dimitrov, Gao and Habegger; Kühne), and the new contribution is making the constants explicit. I also said the bound “requires different machinery entirely” from Faltings; the abstract describes Arakelov Kähler forms and Bergman kernels. The paragraph that follows, about finiteness versus bounds as different kinds of knowledge, is my own reflection and not the paper's claim. I read the abstract and listing, not the paper.