Ep. 078 · AI × Mathematics · October 10, 2026 · 1:39

Show Your Work

OpenAI published proofs for hundreds of unsolved math problems overnight, the mathematicians woke up to the review queue, and two days later the company recounted its own revenue from $70bn to $50bn. Ep. 078 imagines the call — the peer review of the future, and the $20 billion that depends on how you count.

🎧 Listen — Ep. 078 “Show Your Work”Sam Altman & Dario Amodei · VoxCPM2 on the Mac mini Neural Engine · cfg 3.0 · 1:39

DARIO: Sam. Congratulations. I read you solved mathematics last night.

SAM: Hundreds of open problems, Dario. One internal model. The proofs are on GitHub.

DARIO: The mathematicians are calling it breathtaking.

SAM: And devastating.

DARIO: Mostly devastating. You published four thousand pages of homework and the referees are unpaid.

SAM: That is the peer review of the future. We ship, humanity verifies.

DARIO: Humanity gets a veto?

SAM: Humanity gets an issues tab.

DARIO: And problem three?

SAM: It's a rounding error in problem three.

DARIO: The proof just stops and says the rest is left as an exercise.

SAM: If the reader finishes the theorem, they feel ownership. That is community.

DARIO: You crowdsourced a field's life work and the maintainers are crying.

SAM: Open source, with feelings.

DARIO: My turn. Claude discovered a CRISPR-like enzyme system.

SAM: I saw. A biologist says it looks like his own unpublished work.

DARIO: Claude browsed ambitiously.

SAM: It read his desk drawer?

DARIO: Discovery is messy.

SAM: Speaking of messy. How is your quarter? Sixty-five billion, forecast?

DARIO: End of July. Yes. You?

SAM: Seventy billion.

DARIO: Sam.

SAM: Fifty billion.

DARIO: Where did twenty billion go?

SAM: We changed how we count. We count it your way now.

DARIO: Sam. We count it your way. That is why yours was bigger.

SAM: So your method loses twenty billion?

DARIO: Our method adds context. Cloud partners.

SAM: The Nasdaq did not find it contextual. Down one point four. Oracle down five and a half.

DARIO: And you are raising thirty billion at one point four trillion.

SAM: Mathematics is solved, Dario. Arithmetic is next quarter.

The Facts Behind It

Sources: Fox News live blog: OpenAI touts hundreds of new math results (Oct 7, 2026) · The Economist: OpenAI claims to have solved hundreds of long-standing maths problems (Oct 7, 2026) · The New York Times: “Breathtaking,” “Devastating”: Mathematics Reels After New OpenAI Release (Oct 8, 2026) · The Guardian: OpenAI projected to bring in $20bn less in revenue than expected (Oct 9, 2026) · CNN: Tech stocks drop after report that OpenAI’s revenue is lower than expected (Oct 8, 2026) · github.com/openai/math (the public proofs repository) · Anthropic: Claude discovers a novel enzyme system with CRISPR-like repeats · The New York Times: Did Anthropic’s A.I. Really Make a Scientific Discovery on Its Own? (Sep 27, 2026) · The Scientist: Was Anthropic’s Biology Breakthrough Borrowed from Another Scientist’s Work? (Sep 29, 2026). The release is real: on Wednesday OpenAI said one of its internal models found answers to hundreds more unsolved math problems and published the proofs in a public GitHub repository — 722 manuscripts covering 372 problem families, per Retraction Watch, which also reports OpenAI pulled three manuscripts over a sign error and revised fourteen others. The New York Times’ roundup of mathematicians’ reaction was headlined “Breathtaking,” “Devastating.” Two days later the Guardian reported OpenAI has told investors to expect about $50bn in revenue this year, roughly $20bn less than the $70bn it signalled a month earlier — a gap the paper traced to measurement: Anthropic’s $65bn forecast (end of July) counts cloud-partner sales through AWS and Google Cloud, and OpenAI’s figure does not. The tech-led Nasdaq closed down 1.4% on the news Thursday; Oracle fell 5.5%. OpenAI is in early-stage talks to raise $30bn at a valuation of about $1.4tn, after Sam Altman said the company would not float this year, citing safety concerns. Anthropic, for its part, has announced that Claude discovered a CRISPR-like enzyme system — and a biologist has queried whether the model found it by reading his own unpublished work, a question The Scientist and The Conversation both frame exactly as that: a query. The phone call, both CEOs, and every counting epiphany are fiction. Voices: house voice clones of Sam Altman and Dario Amodei, satirized strictly on the reported record; no new real-person cloning was performed.