Ben Recht: OpenAI's 'faster FFT' is a clever exploit, not a math breakthrough
beenwrekt · x · 2026-10-12
Berkeley researcher Ben Recht examines the algorithm in OpenAI's math repository that claims a faster Fourier transform, and concludes it is "less a breakthrough and more a malicious exploit" of mathematical traditions.
Key points:
- The classic FFT (Cooley & Tukey, 1965) recursively splits a length-n transform into two n/2 transforms with linear recombination, iterating log2(n) times, cutting complexity from naive O(n²) to O(n log n).
- Recht walks through OpenAI's claims and implementation, arguing its practical gains are overstated and the technique leans on gaming established results rather than genuine algorithmic innovation.
- The post includes tutorial material on signal processing and computational complexity for readers who need a refresher.
Related event: OpenAI's Faster FFT Claim Sparks Academic Backlash(4 posts)→
More from Research
- Empiricism Comes for Mathematics: AI Proofs Break a Centuries-Old Alignment — manoelribeiro · 2026-10-13
- Meta FAIR's h1: a minimal Lean feedback loop beats hand-designed theorem-proving workflows, hits SOTA — burny_tech · 2026-10-13
- MIRAGE anchors feed-forward vision to LiDAR-IMU motion for metric SLAM and Gaussian reconstruction — zhenjun_zhao · 2026-10-13
- Zebra3D: hybrid sparse-linear attention speeds 3D foundation models up to 7.1x — zhenjun_zhao · 2026-10-13
- CompactSplat cuts 3D Gaussian counts 5x with texture-adaptive allocation for feedforward scenes — zhenjun_zhao · 2026-10-13
- Confidence-aware densification beats 3DGS-MCMC across all budgets, recovering 0.84dB of 1.05dB PSNR gap — zhenjun_zhao · 2026-10-13