October 2026
An OpenAI preprint proves that a fixed Hodge-generic curve in the moduli space \(A_2\) contains only finitely many rational points whose associated abelian surfaces are isogenous to a product of elliptic curves with at least one complex-multiplication factor. The authors establish this component of the Zilber–Pink conjecture by combining André–Oort in the case where both factors have CM with a Galois-orbit bound and a counting theorem in the mixed case. The result allows arbitrary isogeny degrees and imposes no additional boundary or reduction hypothesis.
An OpenAI preprint proves pro-modularity at 2 for a broad class of two-dimensional Galois representations, without requiring them to be de Rham at 2 or to have irreducible residual representations. The authors’ claim is about occurrence in a completed Hecke algebra spanning modular forms of many weights, not realization by a single classical form. Their proof constructs one auxiliary representation with stronger local properties to establish classicality, then uses a dimension comparison to transfer Hecke support to the wider family.
An OpenAI preprint argues that complete packets of lattices arising from primitive totally real quartic fields become equidistributed as the discriminants of their orders grow. Its theorem covers arbitrary orders and averages over all the packet’s torus orbits, rather than asserting equidistribution for individual lattices or selected ideals. The proof rules out a specific non-Haar limit—a decomposition into two coordinate blocks—before concluding that the packet measures converge to Haar probability.
An OpenAI manuscript states that a restricted quotient of a field’s absolute Galois group, together with its commutator bracket, can recover the field’s perfect closure and constant field when the relative transcendence degree is at least two. The Bogomolov–Pop reconstruction theorem allows one global \(\ell\)-adic unit scaling in the group data and, in positive characteristic, identifies field isomorphisms up to powers of Frobenius. The manuscript’s central technical step is a uniform bound on the support of images of a pencil of functions, which lets the authors recover the geometric data needed for reconstruction.
An OpenAI manuscript on Milne’s rationality conjecture proves a specific cross-theory agreement for abelian varieties with good reduction: for a fixed rational Hodge class and fixed divisors on the reduction, a normalized trace is the same rational number in the specified \(\ell\)-adic and crystalline cohomologies. The proof does not require new divisor classes on the reduction to lift to the original variety; instead, it transports the class through a CM variety and, in one case, constructs a new lift that preserves the relevant divisor data.
An OpenAI manuscript on the weak inhomogeneous Duffin–Schaeffer conjecture gives a sufficient condition for infinitely many shifted rational-approximation hits for almost every real number. For a fixed shift \(\gamma\) and non-negative tolerances \(\psi(q)\), the authors prove the result when \(\sum_q (\phi(q)/q)\psi(q)\) diverges, without requiring the approximating numerators to be coprime to their denominators. The proof’s main challenge is controlling the structured overlap among approximation intervals.
An OpenAI preprint proves that infinitely many primes p have p−1 squarefree with an even number of distinct prime factors. Its proof uses weighted candidates and analytic estimates to show that enough survive as primes after composites are removed; it does not claim that half of all primes have this property. Perplexity Computer adapted the preprint into an animated video.
An OpenAI preprint on the joint Dickman law argues that the normalized sizes of the largest prime factors of consecutive integers become statistically independent when integers are counted uniformly over an expanding range. In the limit, the chance that both largest prime factors fall below specified power thresholds is the product of the two familiar Dickman probabilities. The result also implies that each neighbor has the larger largest prime factor half the time, in natural density.
An OpenAI preprint on Jacobsthal’s function claims that if a modulus has \(k\) distinct prime factors, every interval of length \(Ck^2/(\log\log 3k)^2\) contains an integer coprime to it, for an absolute constant \(C\). The manuscript’s main argument establishes a positive count of survivors after small-prime exclusions, then shows that larger prime factors cannot eliminate them all. It does not give a numerical value for \(C\) or determine the bound’s optimal growth.
The manuscript claims that, under its stated hypotheses, an odd, regular, two-dimensional Galois representation over the rationals with two-adic coefficients comes from a classical cuspidal eigenform, up to a Tate twist, without a restriction on its image modulo 2. It divides the proof into three stages: showing that the target eigensystem occurs in completed quaternionic forms, establishing classicality through a separate local argument at 2, and descending the resulting classical packet to the rationals.
An OpenAI preprint proves the existence of a nonzero perverse Hecke eigensheaf for \(SL_n\) bundles with flags at several marked points, when the associated \(PGL_n\) local system has Zariski-dense image and tame, unipotent monodromy at each marking. The theorem permits any unipotent local shape—not just regular unipotent monodromy—and preserves the full tensor and fusion structure of the Hecke operations. It requires no Frobenius structure.
An OpenAI preprint claims that every Dirichlet L-function, including the Riemann zeta function, has no zeros strictly to the right of Re s = 11/12—a fixed boundary that falls short of the Riemann hypothesis. Its proof, as presented in the manuscript, first bounds cancellation in a weighted sum over ideals in the Eisenstein integers, then uses a Mellin transform to turn that bound into a zero-free region. The key step is a deep mean-square estimate for a family of character sums; the paper’s account of that estimate is a roadmap rather than a full proof.
An OpenAI preprint proves that a reduced, irreducible, closed curve defined over the algebraic numbers and Hodge generic in the Siegel threefold contains only finitely many points whose abelian surfaces have an indefinite quaternion division algebra as their full rational endomorphism algebra. The finiteness holds across all such algebras, rather than separately for each fixed type. The proof establishes a height bound independent of the algebra, then rules out unbounded degrees by showing they would force the curve’s period image into a locus too small to contain it.
An OpenAI preprint proves that, in each fixed prime degree \(n \ge 5\), the volume-weighted measures on complete packets of arithmetic torus orbits converge to Haar probability as the discriminants of their multiplier orders grow. The result allows arbitrary local module types, including noninvertible modules, but does not claim equidistribution for an individual orbit or cover composite degrees.
An OpenAI preprint proves a conjecture about the prime factors of \(p-1\): for a uniformly sampled large prime \(p\), any fixed number of the largest prime-factor occurrences, measured by their shares of \(\log(p-1)\), converge jointly to the leading pieces of a Poisson–Dirichlet partition with parameter one. The result is a limit in distribution, not a claim that each individual prime has a typical factorization. Its proof combines estimates for factor counts in arithmetic progressions with a size-biased sampling argument that yields the stick-breaking law.
An OpenAI preprint claims that a curve in the moduli space of principally polarized abelian surfaces contains only finitely many points isogenous to the square of a non-CM elliptic curve, even when the elliptic curve and isogeny vary. The result applies to closed, irreducible curves defined over the algebraic numbers that are Hodge generic, with no boundary condition. In the proof strategy, comparisons at two auxiliary prime levels constrain the targets; height and orbit-size estimates then turn those local constraints into finiteness.
A September 2026 manuscript claims to prove Goldfeld’s density prediction for every elliptic curve over the rationals: among its quadratic twists, half have analytic rank zero and half rank one. Its key step is a converse linking arithmetic information to analytic rank: when a twist’s full two-power Selmer group has corank zero or one, the authors claim its analytic rank matches that corank. They then use a distribution theorem for Selmer coranks to derive the predicted densities.
An October 2026 manuscript claims that the Margulis–Platonov conjecture holds for absolutely almost simple, simply connected groups over global function fields, including fields of characteristic two. It argues that every normal subgroup not contained in the center is determined by its image at the finitely many places where the group is anisotropic: specifically, it is the inverse image of an open normal subgroup of the product of the corresponding local groups. The proof reduces the claim to showing that a finite gap between a subgroup generated by powers and its local closure must vanish.
An OpenAI preprint, *Étale covers with a prescribed exterior sheet*, constructs a finite étale cover of a \(p\)-adic curve that is trivial on one sheet over the exterior of fixed disks, while every component above each disk has degree divisible by a chosen prime \(p\). The authors’ argument first builds cycles above the disks, then uses a group-theoretic quotient to enforce the degree condition on every disk component without losing the exterior sheet.
A September 2026 manuscript, *The Additive Indecomposability of the Primes*, argues that no two fixed sets of nonnegative integers, each containing at least two elements, can have all their pairwise sums match the primes apart from finitely many exceptions. The authors’ claim is stronger than ruling out an exact representation: any such addition table must produce infinitely many composites or miss infinitely many primes. Their proof combines constraints on residues modulo primes with estimates linking the summands’ arithmetic structure to the requirement that every sufficiently large prime be represented.
The preprint *Constructible tame Hecke eigensheaves in positive characteristic* proves an existence result for nonzero, constructible Hecke eigensheaves associated with Zariski-dense local systems and regular unipotent tame monodromy at a puncture. Its construction specializes first from characteristic zero and then to positive characteristic, under stated hypotheses on the curve, group and characteristic. To address the risk that removing auxiliary flag data erases cohomology, the article makes explicit a local coinvariant argument for nonvanishing; it does not develop the broader globalization of the trace identity behind that argument.
OpenAI’s September 2026 manuscript on the Margulis–Platonov conjecture claims that, for absolutely almost simple, simply connected groups over number fields, every non-central normal subgroup of the rational points is the exact preimage of an open normal subgroup in a finite product of compact local groups. The result does not require the normal subgroup to be closed or finitely generated. The proof reduces the classification to a finite gap between a subgroup and the closure of its local image, then rules out that gap through arithmetic and finite-group arguments.
Pollen Robotics and Hugging Face introduce Reachy Mini, an open-source desktop robot, through a music video built around the machine’s own recorded movements. The video’s case for the small robot rests on its expressive head and antenna motions, visible construction and simple prompt-and-response interaction; it also extends the open-source invitation to the film itself, which viewers can scrub through and fork.
OpenAI’s mathematical results illustrate the rapid progress John Coogan sees in tasks that can be carried out and checked in digital environments, but he argues that technical capability does not settle the broader work of building a company. That work still involves judgment, hiring, dealmaking and execution. In the discussion of SpaceX’s proposed $40 billion chip financing, Jordi Hays argued that investors must assess the risks and valuation themselves, even if Elon Musk’s track record earns him some latitude.