What happened
Axiom’s November 6, 2026 announcement describes a Lean formalization of the four classical Weil conjectures. These concern the zeta functions that encode point counts for varieties over finite fields: their rationality, a functional equation, a relationship to topological Betti numbers under good reduction, and the predicted magnitudes of their roots. Axiom says the formalization includes the comparison with Betti numbers and packages the point-counting conclusions as a theorem that later Lean proofs can invoke.
The development builds on Mathlib, the Lean community’s mathematics library, and Axiom’s formalization of the Stacks Project. Axiom says the new work uses the same formal definitions as its Stacks formalization, so earlier proofs can be used directly rather than reconstructed. The company identifies *Étale Cohomology and the Weil Conjecture* by Eberhard Freitag and Reinhardt Kiehl as the principal reference. It says the decisive weight estimate comes from its Stacks formalization, following Gérard Laumon’s Fourier-transform route. The result therefore does not reproduce Pierre Deligne’s 1974 proof line by line.
Axiom says Lean checks whether the stated conclusions follow from the definitions and assumptions. The repository reports a proof with no sorry placeholders and no added axioms beyond Lean’s standard logical foundations. The announcement also describes technical refinements: relevant characteristic polynomials have integer coefficients and are independent of the auxiliary prime used to construct cohomology. It says the central eigenvalue and point-counting results apply to smooth proper varieties, extending beyond projective varieties.
The announcement emphasizes that this is a reusable formal result, not a new mathematical proof of the conjectures in the ordinary sense. It encodes established mathematics in a form that Lean can check and subsequent formal developments can depend on. Axiom names contributors from its team and the wider mathematical and Lean communities, including lead contributor Tony Yue Yu.
What the archive adds
New Runtime’s earlier coverage described Axiom’s formalization of the Stacks Project on November 5, 2026, as a reusable Lean library of about 16 million lines. This is a new event on that known thread: Axiom says the Stacks foundations now support a formalization of the Weil conjectures.
Trend signal
This strengthens the case for reusable formal artifacts as a foundation for later research rather than one-off demonstrations. It is relevant to New Runtime’s earlier coverage of AI proof systems and machine-checked mathematical results, though Axiom’s announcement is about a formalization project, not a benchmark of model ability. This does not yet fit a pattern New Runtime tracks.
Why now
Axiom presents the formalization as a step made possible by its earlier Stacks work: the foundations and decisive weight estimate are already available in Lean. The announcement points to a next application—formalizing the connection between modular forms and cohomology needed to derive Ramanujan’s bound—but says that work is not part of this release.
Connections to other events
New Runtime’s earlier coverage of the Stacks formalization is the direct technical connection: Axiom says the Weil development uses its shared definitions and proven results. The new result also follows the broader direction represented in New Runtime’s earlier coverage of Lean proof systems, where machine-checked mathematics is treated as an inspectable artifact rather than only a claim about what a model can solve.
Why it matters
For teams building formal mathematics workflows, the important unit is not just a verified theorem but the dependency chain beneath it. If the foundations are already encoded and reusable, a new project can start from a checked result instead of rebuilding definitions and proofs. That makes library structure and compatibility consequential: Axiom says its Weil development uses the same formal definitions as its Stacks work and standard Mathlib definitions.
The release also illustrates the gap between formal checking and project assurance. Lean can verify that a proof follows from its definitions and assumptions, but that does not by itself establish that the formalization captures the intended mathematics or that the project is easy to maintain. Human review, clear scope, and a legible dependency graph remain important before other teams rely on a substantial new library.
Limits and uncertainty
Axiom reports that the repository has not yet been reviewed by humans. The announcement does not provide an independent audit of the formalization, a measure of its maintenance burden, or evidence that other projects have adopted it. The formalized result covers the Weil conjectures described by Axiom, but the proposed Ramanujan application is explicitly outside this release. Its proof will still need a formal connection between modular forms and cohomology.
Reader takeaway
Axiom has turned the Weil conjectures into reusable, machine-checked Lean results built on its Stacks formalization. The practical significance is the accumulated foundation future formal proofs can use—not a claim that the underlying mathematics is new or that the project has already received independent human review.