Scholars Clash Over OpenAI's Faster FFT Claim

After OpenAI published an algorithm in its math repository claiming a faster Fourier transform (FFT), a heated academic debate erupted on X. Berkeley scholar Ben Recht (beenwrekt) analyzed the algorithm on his blog arg min, concluding it is "less a mathematical breakthrough than a malicious exploitation of mathematical tradition." He joked that nobody reads AI papers anyway: "I did what mathematicians worldwide do with AI papers: fed it back to ChatGPT and asked what it thought." Critic JoshPurtell relayed the take that AI excels at automating humanity's worst academic instincts—amplifying work no one will read.

Confirmed

OpenAI released an algorithm claiming a faster FFT, originating from its math repository (per Ben Recht's blog analysis).
FFT's classical foundations were laid by Cooley and predecessors; Recht argues the new result merely exploits the notation and definitions of mathematical tradition (key point from Recht's blog).
The debate unfolded publicly on X, with beenwrekt's biting commentary amplified by vikhyatk and many others.

Points of Contention

Yoav Goldberg offered a dissenting view: OpenAI merely published a theoretical result consistent with existing conclusions in theoretical computer science (TCS), so accusing it of "abusing TCS definitions" is odd; any criticism should target the TCS field itself, not OpenAI's specific result.
Goldberg added that the earlier sub-cubic matrix multiplication result was equally impractical, suggesting this is better framed as a fair critique of the TCS field.

Why It Matters

The dispute reflects a deeper fight over how to evaluate AI-generated "mathematical results": asymptotic improvements may not be practical, and the community remains highly sensitive to the gap between marketed claims of "AI doing math" and actual theoretical contributions.

2026-10-12 ~ 2026-10-13 · 6 related posts

Full story(2 episodes)→

Episode 1: OpenAI's New Multiplication Algorithm Sparks Debate Over Practical Value(2026-10-12, 2 posts)
Episode 2: Scholars Clash Over OpenAI's Faster FFT Claim(2026-10-12, 6 posts)

Primary sources

Ben Recht: OpenAI's 'faster FFT' is a clever exploit, not a math breakthrough — beenwrekt · 2026-10-12 22:28 (UTC+08:00)
OpenAI FFT paper called 'exploit of math traditions'; Goldberg pushes back — yoavgo · 2026-10-12 23:23 (UTC+08:00)
OpenAI's FFT result sparks debate: math breakthrough or automated bad academic habits? — JoshPurtell · 2026-10-13 01:18 (UTC+08:00)

[source] Ben Recht: OpenAI's 'faster FFT' is a clever exploit, not a math breakthrough — beenwrekt · 2026-10-12
[source] OpenAI FFT paper called 'exploit of math traditions'; Goldberg pushes back — yoavgo · 2026-10-12
[source] OpenAI's FFT result sparks debate: math breakthrough or automated bad academic habits? — JoshPurtell · 2026-10-13
Mathematicians mock OpenAI's FFT paper: 'No one at OpenAI read it either' — ZeroStateReflex · 2026-10-13
Ben Recht: OpenAI's FFT algorithm is a malicious exploit, not a breakthrough — AlexTensor · 2026-10-13

1 near-duplicate retellings: yoavgo