Joke thread pokes at claimed sub-n log n results: it's actually n(log n)^0.5477…
burny_tech · x · 2026-10-07
A retweeted thread lists theoretical CS results attributed to OpenAI's internal frontier model: improving 3sum to O(n^1.9992), integer multiplication to roughly O(n log n^(1e-55)), and Fourier transform gains at n log n^(1e-13).
The quoter follows up with a joke: a result billed as "below n log n" turns out to be n·(log n)^0.54777… with a long trail of 9s — an asymptotic improvement barely distinguishable from the original bound.
Related event: OpenAI's 'Below n log n' Claim Called Out as Wordplay(2 posts)→
More from Fun
- Reality check: a 1-minute laptop simulation gets within 0.00000007% of the 11-squares optimum — AaronBergman18 · 2026-10-07
- "Do (x) for agents": fill in the blank and YC funds it, says founder — gaganghotra_ · 2026-10-07
- Making more money just to afford more OpenAI and Claude subscriptions — AaronBergman18 · 2026-10-07
- "How it feels to have never learned to code" meme goes viral in vibe coding circles — MattGarciaEth · 2026-10-07
- Ben Affleck explains how Transformer AI differs from classical computing — rohanpaul_ai · 2026-10-07
- Microwave bros once claimed restaurants were cooked — a jab at today's AI hype — kylegawley · 2026-10-07