Erdős Problem #888 — What is the size of the largest A⊆{1,…,n}A\subseteq \{1,\ldots,n\} such that if a≤b≤c≤d∈Aa\leq b\leq c\leq d\in A are such that abcdabcd is a square then ad=bcad=bc?

About 28 years old · traced to

What is the size of the largest A⊆{1,…,n}A\subseteq \{1,\ldots,n\} such that if a≤b≤c≤d∈Aa\leq b\leq c\leq d\in A are such that abcdabcd is a square then ad=bcad=bc?

References

Progress summary

Refreshed
Claimed solved

A proposed AI-generated argument claims the conjecture’s predicted growth rate, but no independent verification has established it.

The problem asks for the largest admissible subset of [1,n][1,n], conjectured to have order nlog⁡log⁡n/log⁡nn\log\log n/\log n. The formal Lean statement remains incomplete, with the main theorem containing sorry.

Known results

  • Sárközy proved an o(n)o(n) upper bound; Tao later supplied a proof.
  • The primes give a lower bound of order n/log⁡nn/\log n.
  • The primes together with squarefree semiprimes give (1+o(1))nlog⁡log⁡n/log⁡n(1+o(1))n\log\log n/\log n.

May 2026 claimed upper bound

The site attributes to GPT-5.5 Pro, prompted by Chojecki, a colored bipartite-graph argument claiming the matching upper bound ∣A∣≪nlog⁡log⁡n/log⁡n|A|\ll n\log\log n/\log n. If correct, this settles the conjectured order, but the proof remains unverified and no independent mathematical publication or check is recorded.

Current status (as of September 2026): The lower construction and earlier o(n)o(n) upper bound are established, while the matching upper bound is claimed but remains unverified; the problem is not formally resolved.

Sources

Solutions 0

No solutions have been posted yet.