VibeMathedMath problems solved by AI
All problems

Erdős Problem #671

Erdős problem #671 · erdosproblems.com/671

For triangular arrays of nodes ain[1,1]a_i^n\in[-1,1] let Lnf\mathcal{L}^nf be the Lagrange interpolation polynomials of a continuous ff, with fundamental polynomials pinp_i^n. Is there a choice of nodes such that for every continuous ff there is some xx where lim supnipin(x)=\limsup_n \sum_i\lvert p_{i}^n(x)\rvert=\infty and yet Lnf(x)f(x)\mathcal{L}^nf(x) \to f(x)? Is there a choice with lim supnipin(x)=\limsup_n \sum_i\lvert p_{i}^n(x)\rvert=\infty for every xx, yet for every continuous ff some xx has Lnf(x)f(x)\mathcal{L}^nf(x)\to f(x)? Both questions are claimed resolved in the affirmative.

Result
Proved(see note)
Status
Candidate (review pending)
AI contribution
AI-discovered
Method
Construction
Field
Analysis, Interpolation
Posed by
Paul Erdős
Year posed
1982
Years open
44y
Solved
2026-06-22
Model
GPT-5.5 Pro, Codex
Vendor
OpenAI
Collaborators
Liam Price
Verification
Lean-verified
Publication
Announced
Significance
10 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

Both parts claimed answered affirmatively, with a Lean formalization; two proof claims are filed on erdosproblems.com but the problem is still listed open

What the AI did

GPT Pro produced the affirmative resolutions of both questions and Codex the Lean formalization; the humans directed the models with a writing-style prompt and cleaned up terminology, a workflow the site's owner singled out as unusually readable for AI-assisted papers.

Verification

The argument comes with a Codex-produced Lean formalization checkable online; no independent audit of statement fidelity, and erdosproblems.com still lists the problem open with the claims filed.

Sources

erdosproblems.com/671

Discussion