Normal orders of the Erdős chain functions h(n) and H(n)

David Turturean

Turturean’s write-up claims h(n)=(1/2+o(1))log* n and H(n)=(1+o(1))log* n for almost all n, so H(n)/h(n) tends to 2. A GPT-5.5 Pro subset-product successor lemma supports the composite-divisor lower bound.

Sources for this record

What was already known

Erdős asked whether the ratio diverges. Shengrong Wu earlier posted an AI-generated proof of two-sided constant-factor iterated-logarithm bounds; the exact 1/2 and 1 constants are the claimed further advance.

AI involvement AI-led

The repository credits GPT-5.5 Pro with the result statement, sketch and successor lemma, and Claude with manuscript assembly and Lean work. Turturean directed, assembled and audited the project.

Multiple original sources explicitly assign the central proof sketch, successor lemma and formalization to models; human assembly and checking remain material.

Importance Notable

The claimed constants sharpen the earlier bounds and give a finite limiting ratio for Erdős 696.

Limitations

The full argument also relies on prime-chain estimates and imported analytic number theory. Turturean reports auditing the imported inputs, but no completed end-to-end check of the final theorem is documented.

Technical verification Partial

The repository reports a sorry-free Lean formalization, but its main theorem depends on three external analytic-number-theory axioms. The three inputs remain assumptions in the formal theorem, and the source comparison does not establish a complete unconditional check.

Formal checking

Approximately 19,200-line sorry-free Lean development with three separately listed external axioms

Checker
Lean 4 compiler/kernel in Turturean's reported build
Scope
Formal Erdos696 main result conditional on the stated Siegel–Walfisz, Brun–Titchmarsh and Mertens inputs
Basis
Reported by the linked checker
Evidence source

Checking a formal statement does not establish that it matches the original problem.

Human understanding Few

Default for missing evidence of full human grasp: Turturean describes trying to understand and rederive the proof, but no completed human explanation of the full exact-constant mechanism is documented in the sources reviewed.

Correction and publication history

2026-10-18 · publication

The claimed constants sharpen the earlier bounds and give a finite limiting ratio for Erdős 696.

Read this published revision