Research notes on every problem, and a library of the papers behind them. Built from the open erdos repository.
Updated
The claim. For every , the -uniform hypergraph on the
vertices whose edges are the triples with
one vertex in each part together with the triple inside
the first part has at least edges, no four vertices spanning three edges
and no five vertices spanning seven. So the statement of
Problem 794 fails for every
, and the answer to the question as written is no. The result is the
theorem construction_properties of the Lean file
src/v4.24.0/ErdosProblems/Erdos794c.lean in Boris Alexeev's lean-proofs
repository (plby/lean-proofs), linked above at the commit of 15 September 2026
(Lean and Mathlib v4.24.0). Alexeev's comment of 5 February 2026 on the site's
discussion thread, which announces the Lean check of Harris's example, adds in a
later update (the file it links was first committed on 6 February 2026) that
Aristotle proved the result by itself without a proof provided, using
essentially the same example, and links this file. The file carries no
informal-author header: it states the construction and proves its properties in
Lean, so it is an independent proof by an AI system, published by Alexeev, and
has its own page; the Lean checks of Harris's example are links on
Harris's page.
The file's construction_properties n (h : 3 ≤ n) asserts, for
construction n (the union of tripartite_edges n and extra_edges n,
the latter empty when ), that every edge has three vertices, that the
edge count is at least n^3 + 1, and that there is no vertex set of size
spanning at least edges and none of size spanning at least ;
its closing comment records #print axioms as propext, Classical.choice
and Quot.sound. The theorem states the construction's properties for each
and does not itself negate the problem's quantified statement; the
negation follows by instantiating any . The hypothesis is
needed, since a part with fewer than three vertices holds no triple.
Depends on. Nothing in this wiki: the construction and its verification are the file's own.
Acceptance. None recorded. The corpus has not built or audited the file and
holds no statement-fidelity audit of it, so formalized is not listed; no
outside reviewer is recorded for this file, and the site's label DISPROVED
(LEAN) and the formal-conjectures formal_proof attribute both name the Lean
check of Harris's example, not this file. No journal publication exists and none
is expected. The claim stays claimed; the problem's standing rests on the
accepted page of Harris.