VibeMathedMath problems solved by AI

Cutoff for Kac's Walk on the Sphere

The discrete-time Kac walk on Sn1S^{n-1} started from a coordinate vector exhibits total variation cutoff at CBRWnlognC_{\mathrm{BRW}} n \log n, where CBRW3.8916C_{\mathrm{BRW}} \approx 3.8916 is set by the speed of the leftmost particle in a branching random walk. The cutoff is therefore not at the conjectured 2nlogn2n\log n.

Result
Proved(see note)
Status
Resolved
AI contribution
AI co-developed
Method
Argument
Field
Markov chains
Posed by
Mark Kac
Year posed
1956
Years open
70y
Solved
2026-07-15
Model
ChatGPT
Vendor
OpenAI
Collaborators
Vishesh Jain, Clayton Mizgerd
Verification
Unreviewed
Publication
Preprint
Significance
20 / 100
Disclosed cost
Wikipedia
No dedicated article

What was actually shown

the conjectured cutoff location of 2n log n is wrong

What the AI did

The acknowledgement says the authors used ChatGPT extensively, at the level of a co-author, for brainstorming, help with proofs, literature review, checking for mistakes, writing code, and preparing the manuscript. It does not separate which parts came from where.

Verification

arXiv preprint; not yet peer-reviewed.

Source

arXiv:2607.13401 - Total variation cutoff for Kac's walk on the sphere

Discussion