Erdős Problem #1148
Editorial depth: Authored explanatory DRAFT — NOT editorially accepted.
Snapshot classification: proved (not automatically the live result).
Subject areas: number theory.
Read this status correctly
The status above is reproduced from GCL's immutable external-catalogue snapshot. It is not a new claim of current mathematical resolution, a source-semantic review, an independent proof check, or GCL certification. A published proof and a GCL replay are separate events.
UNACCEPTED DRAFT — mathematical editorial review pending
This material is a provisional, source-qualified manuscript published for line-by-line scrutiny and real HTML/LaTeX testing. It has NOT received separate GCL editorial acceptance or independent mathematical certification. Its presence on this website is not evidence of a solution, reviewed coverage, or final publication.
Review the underlying draft on GitHub.
The question in plain language
Erdős Problem #1148 concerns the following mathematical question and its exact quantifiers:
For every sufficiently large positive integer n, find INTEGER x,y,z satisfying n=x²+y²-z² with max(x²,y²,z²)<=n. Signs do not change the squares; x,y,z may be chosen nonnegative. This is an EVENTUAL representation theorem, not assertion that every n is representable. Important that component squares separately stay <=n, not merely absolute value bound of x²+y²-z².
A small example
An independent exact perfect-square set search verifies for every n<=300 which representations exist and that 6563 has NO representation with all x²,y²,z²<=6563; 6564 DOES have one. This is a bounded witness and not proof of a finite exceptional set or the full ELMV12 mechanism. Fixture uses identity n+z²=x²+y² with each square<=n.
What is known
https://www.erdosproblems.com/1148 PROVED (LEAN) as of May 28 2026, citing Chojecki and GPT-5.4 Pro; deduction using a form of Duke's theorem via Einsiedler–Lindenstrauss–Michel–Venkatesh [ELMV12]. Original Vardi [Va99,1.25] notes largest known failure 6563. Website reported Lean verification is NOT yet source-pinned/kernel-replayed by GCL. The current maintained catalogue postdates frozen initial Lean theorem with local sorry.
GCL's current state
Authenticated original-issue source return established the bounded source and finite-evidence interface. The number \(6563\) illustrates a genuine finite obstruction; the theorem requires eventual representability beyond a finite exceptional set.
All general results cited here remain external until primary-source review.
What remains
ERDOS-1148-CHOJECKI-DUKE-LEAN-001: read full Chojecki/2026 note, source-lock precise Duke equidistribution input (ELMV12) and reduction, pin exact external Lean commit and check theorem, sorry and transitive axioms, and replay representative finite gaps including 6563.
Primary analytic number theory proof + external Lean dependencies not GCL-reviewed. External proved status, no GCL new proof.
Sources and claim boundaries
https://www.erdosproblems.com/1148 ; https://www.erdosproblems.com/forum/thread/1148 ; Vardi 1999; [ELMV12], Chojecki/GPT-5.4 Pro 2026; phase17 exact enumeration.
This is a source-bound, unaccepted editorial draft. Neither protected cohort admission nor proof or Lean certification is implied.
Catalogue and evidence provenance
- Pinned catalogue source: teorth/erdosproblems at
536408716724. - GCL issue: #2419 — protected issue/assignment identity; its labels and results are live, not reproduced automatically here.
- GCL triage: OUTSIDE_STRICT_OPEN_TRIAGE.
- Formal-statement coverage: A provider statement was included in the pinned snapshot. Presence does not mean the statement or proof has been checked.
- External mathematical status: consult current problem page and its dated bibliography.