AI-led Lean proof of the Gurvich–Naumova partition conjecture

For every n≥1, the integers 1 through n can be partitioned into nonempty blocks of at most three distinct integers, with each block summing to a power of three.

Sources for this record

What was already known

Gurvich and Naumova proved partial cases and reduced the possible counterexamples, but the full base-three partition conjecture remained open.

AI involvement AI-led

JD Jones chose and directed the problem. Codex GPT-6 agents developed the proof constructions, finite certificates and Lean implementation, adapting published sequence formulas with credit.

The disclosure credits Codex agents with the substantial proof constructions and searches while Jones supplied direction and the target.

Importance Notable

Proves the full base-three existence statement for every n.

Limitations

The result does not cover other bases, uniqueness or counts of partitions. The source reports Lean, Comparator and proof-replay checks; no independent human mathematical check is documented.

Technical verification Formal (Lean 4/Mathlib, Comparator statement comparison and NanoDa proof replay)

The source reports a complete Lean 4 proof of the exact m=3 statement, finite certificates checked by decide, symbolic unbounded families, a transitive axiom audit and Comparator/independent-kernel replay.

Formal checking

The source reports a complete Lean 4 proof of the exact m=3 statement, finite certificates checked by decide, symbolic unbounded families, a transitive axiom audit and Comparator/independent-kernel replay. This review did not rerun them.

Checker
Lean 4 kernel, Palomar Comparator and NanoDa
Scope
GurvichNaumova.partition_into_powers_of_three for every natural n
Basis
Reported by the linked checker
Evidence source

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

Human understanding Few

The repository gives a proof account, but no person's explanation or independent human review of the central six-family and contraction mechanism is documented; Few is an evidence-gap default.

Correction and publication history