Nutelligence Research · October 2026
Critical site percolation on periodic and quasi-transitive graphs
Independent follow-on research motivated by OpenAI's October 2026 mathematics release, particularly its critical-percolation manuscripts. These statements concern Bernoulli site percolation and must not be conflated with general bond-percolation claims.
This public summary describes the formal claims and evidence. Detailed development files remain internal; publication here does not imply independent peer review or established novelty.
Results and precise scope
Theorem P · Exact Lean proposition verified
For a graph satisfying the formal Zd-periodicity predicate, d ≥ 2, and site critical probability pc < 1, the probability of an infinite open cluster at criticality vanishes at every vertex.
θx(pc) = 0
Lean declaration: SiteBS.Periodic.theoremP_final. MathVerify reports VERIFIED, exact=true, proof=true, with fresh same-kernel replay and authenticated source checks.
Certificate SHA-256: d870d7fdfeb8d18f5de517db9f7eadd1fa856ebe131acbe8ce635b834c09f96f
Source commit: 8767ace492bbbfad469f6e5799d0cbbc16ff1688. GitHub Actions run: 38999327867.
Theorem P′ · Conditional broader-graph result
For connected, locally finite, quasi-transitive graphs with a polynomial bound on ball growth along a sequence of radii and pc < 1, the formal theorem yields zero critical site-percolation probability, assuming the explicitly named TesseraTointonCor24 hypothesis. This hypothesis has not been discharged by the certificate.
Lean declaration: SiteBS.PPrime.theoremPPrime. An exact formal-proposition certificate was issued; the result is conditional and is not an unconditional quasi-transitive theorem.
Theorem S · Specialized lattice family
For dimensions d ≥ 3 and admissible finite sets K, the literal lattice/Cayley graph SiteBS.GK.litGraph d K satisfies the stated vanishing critical site-percolation probability conclusion.
Lean declaration: SiteBS.GK.theoremS_lit. This is a family-specific claim, not a statement about arbitrary quasi-transitive graphs. Its exact formal-proposition certificate is confirmed; downloadable evidence is not yet public.
Proof structure and reproducibility
P: Periodic-graph criticality
The formal target SiteBS.Periodic.theoremP_final applies the project's periodic-graph criticality development to the full proposition with dimension d ≥ 2 and the explicit pc < 1 premise. The exact proof is the pinned Lean declaration and its dependencies; this summary is not a substitute for the source.
P′: Growth and structural reduction
The Lean development first proves subexpGrowth_of_poly_seq, converting polynomial ball growth along unbounded radii into subexponential growth. Next propP94 uses the explicitly assumed Tessera–Tointon structural hypothesis to obtain a finite-index subgroup and nilpotent quotient. Finally theorem71' supplies the criticality conclusion; theoremPPrime_noInfinite expresses the no-infinite-cluster form.
S: Specialized lattice family
The formal proposition SiteBS.GK.theoremS_lit uses the precise definitions of Admissible d K, litGraph d K and the rooted site-percolation event. Its verification does not establish a result for arbitrary lattices.
Exact verification certificates
| Result | SHA-256 certificate fingerprint |
|---|---|
| P | d870d7fdfeb8d18f5de517db9f7eadd1fa856ebe131acbe8ce635b834c09f96f |
| P′ (conditional) | 4b8e23be221b13da16f5094106d249bb82b80b3fbc6a83df8f7eca1b363c7c21 |
| S | c59bc848fb1a2600f51943892420d6d4401267908654ff5755c8f84f18715607 |
All three matching MathVerify records report VERIFIED, exact=true and proof=true for their explicit formal propositions. Full public Lean source bundles, dependency manifests, and downloadable certificates are not yet published. These fingerprints identify internal verification records rather than public certificate downloads.
What formal verification establishes
Lean 4 checks whether a formal proof inhabits its specified proposition. MathVerify orchestrates pinned-source verification and records replay and dependency evidence. Theorem P's certificate concerns the explicit formal proposition. It does not independently establish the correspondence of every informal mathematical interpretation to its Lean definitions, mathematical novelty, or endorsement by OpenAI. An independent second-kernel check is not reported.
Relationship to OpenAI's mathematics release
The research was motivated by OpenAI's critical-percolation manuscripts. The relevant source papers include Critical bond and site percolation on the cubic lattice and No percolation at criticality on quasi-transitive graphs.
The latter addresses bond percolation in a broad graph setting, whereas these Nutelligence formal statements concern site percolation. A proof-level attribution and prior-art review remain necessary before making claims of novel extensions or direct lemma reuse.
Independent scrutiny
We welcome technical review of the periodicity predicate, growth and structural assumptions, proof dependencies, and source-to-formal correspondence. Reproducible public source bundles and further certificates will be linked when cleared for release.