Episode 83 · The Machine's Theorems
Needles, Waves and Dust: Harmonic Analysis Falls

Part 15 of "The Machine's Theorems". Farah and Augustin start with Kakeya's needle, turning a unit segment in every direction using as little room as possible, then climb to Fourier restriction, Bochner–Riesz summability, Sogge's local smoothing conjecture for the wave equation, Falconer's distance sets and Carleson's pointwise convergence problem. We read OpenAI's manuscripts in families 074 (the Kakeya maximal conjecture in 3D, building on Wang and Zahl's 2025 set theorem, and Kakeya sets in 4D), 077 (restriction), 078 (Bochner–Riesz in 3D), 079 (local smoothing in 3D), 073 (Falconer in all dimensions, which is Lean-listed) and 080 (the Carleson Sobolev endpoint). Why the field matters to an engineer who runs FFTs, seismic migration and wave simulations daily.
Listen to the show on your preferred podcast platform.
Inside this episode
- Cold open: a needle in a parking lot
- Kakeya's needle
- Dimension: a finer ruler
- Sticky, plany, grainy: how the set problem fell
- How the maximal proof is organised
- From sets to functions: the maximal conjecture
- What family 074 claims
- Why needles matter: the Fourier connection
- Wave packets with numbers
- The ladder of conjectures
- Restriction: can a sphere's worth of frequencies spike?
- Bochner-Riesz: how much smoothing does Fourier inversion need?
- Halfway: the ladder so far
- Decoupling: listening to the orchestra one section at a time
- Local smoothing: waves are smoother on average
- Falconer: how much dust makes all the distances?
- Why exactly half the dimension?
- A correction to our own brief
- Carleson's problem: when does the wave start where it should?
- L-p norms, said slowly
- What a referee would check first
- What the ladder looks like now
- Why it matters to an engineer
- Closing: the beam turns around
Production credits & AI disclosure
Scripts generated with Claude Opus. Synthetic narration generated with Edge TTS. Reviewed, listened to, and approved by Sankalp Thakur.