“Evaluate all work as if it is AI-assisted.” That is the first of six “very basic concrete suggestions” in a guest post on Terence Tao’s blog, published on 1 November and written by Frank Calegari, a mathematician at the University of Chicago. The post argues that mathematicians should change how they hire, publish and train students now that AI systems produce proofs at scale.
The same day, The New Yorker published a profile of Tao. It reports that in a 9 October talk hosted by the American Institute of Mathematics, he described “math up to 2025” as operating under “proof scarcity”; the magazine says the landscape has changed to one of proof abundance. Every AI proof claim we covered in October, five in all, from Google DeepMind, Microsoft Research, OpenAI, Anthropic and Axiom, cited a check in the Lean proof language, by our count of our own coverage.
Business Pill · Benchmarks
A short explainer on why a benchmark score is not your own result. It teaches the general idea only and says nothing about the guest post or the profile.
The key insight: By our count of our own coverage, all five AI proof claims we reported in October, from Google DeepMind, Microsoft Research, OpenAI, Anthropic and Axiom, cited a check in the Lean proof language, while a guest post on Terence Tao’s blog lists “Is it Lean-verified?” as one question for journals to ask.
Calegari opens with OpenAI. “The release by OpenAI on October 6 of over 700 research papers was a watershed in the history of mathematics,” he writes. He adds that the OpenAI papers, “despite some tremendous ideas — are atrociously written and violate our community expectations of attribution. Some of them are even wrong.”
His case is that the levers are economic. “The tools that are going to be the most effective are going to be economic rather than rhetorical,” he writes, naming hiring and journal acceptance as the tools the profession controls.
The six suggestions, as listed in the post:
“Evaluate all work as if it is AI-assisted.”
“Make hiring much more about personal interactions rather than a paper file.”
“Demand more from journals, and abandon those that do not adapt.”
“Put a greater emphasis on breadth in graduate education, but do not abandon technical skills.”
“Give better seminars.”
“Take more ownership of the results of AI.”
On journals, he asks of each paper: “Is it Lean-verified?” and writes: “We should be producing fewer (and better) papers, not more.” On credit, he writes: “Choosing worthwhile problems may end up mattering more than producing proofs.”
AI-produced or AI-assisted proof claims in FourWeekMBA coverage in October 2026, by date of claim, with the Lean check each source stated. The count is our own; we did not inspect the Lean files.
What Tao Said, per The New Yorker
The profile, by Rivka Galchen, reports that in his 9 October talk, retitled Math 2.0, Tao called “math up to 2025” Math 1.0, which operated under the basic assumption of “proof scarcity.” The magazine writes that he compared the new proof abundance to food abundance, when “you become a bit pickier about what food gets crammed into your mouth.”
The magazine reports that OpenAI released “three hundred and seventy-two additional results” on 6 October, and that OpenAI’s statement mentioned “four thousand problems” had been tried. Calegari’s post gives a different count, over 700 papers; the two sources count different things, and we have not reconciled them with OpenAI’s own statement.
Tao has described AI companies as having “managed to optimize by removing all the humanity out of the process,” according to the profile. It also reports an open letter from an initial group of 25 Fields medallists, Tao among them, saying that “solving problems is only a tool and proxy for achieving the primary goal of conceptual understanding and insight.”
Five October Proof Claims, Each Citing a Lean Check
All five cited a Lean check, by our count; the status is as each source stated it. On 26 October an independent advisory group that OpenAI says it is working with called on it to release the natural-language and Lean outputs behind its physics existence proofs, as we reported.
Why the Labs Are in This Story
Tao told The Observer in October, as quoted on his website, that “maths problems are almost the only thing left for them to train on, so suddenly all the big prizes and goals in our profession are benchmarks for them,” as we reported on 24 October.
As we read the two pieces together, both put the scarce part of mathematics somewhere other than the proof: the guest post writes that “Choosing worthwhile problems may end up mattering more than producing proofs,” and the 25 Fields medallists’ letter, per The New Yorker, calls solving problems “only a tool and proxy.”
As we read the guest post, its proposals target the incentives the profession controls, hiring and journals, which it calls tools that are “economic rather than rhetorical,” rather than the behaviour of the AI labs.
Frank Calegari, guest post on Terence Tao’s blog, 1 November 2026
“Evaluate all work as if it is AI-assisted.”
Three Implications
FOR HIRING COMMITTEES The guest post suggests interviewing postdocs and graduate students in person: “Make hiring much more about personal interactions rather than a paper file.”
FOR JOURNALS The guest post writes: “We should be producing fewer (and better) papers, not more,” and suggests abandoning journals “that do not adapt.”
FOR THE AI LABS In October, the five AI proof claims we covered each cited a Lean check, by our count; on 26 October an independent advisory group that OpenAI says it is working with called on it to release the Lean outputs behind its physics existence proofs.
The framework names one moat this way: “The Verifier Moat — built by RLVR on domain-specific reward signals, building into a new asset class for reasoning-heavy verticals.”
As we read it, mathematics with machine-checkable proofs is the clearest case of that reward signal, which fits Tao’s remark, as we reported in October, that maths problems have become benchmarks for the labs. The framework is the owner’s argument and does not discuss the guest post or the profile.
The guest post is Calegari’s; Tao’s note at the top says only that it is a guest post converted to the blog using AI, and we have not established whether Tao endorses all six suggestions. We did not inspect the Lean files behind any of the five claims; the Lean status is as each source stated it. The 700 and 372 counts come from two different sources and are not reconciled here.
This Week In The AI Supercycle
Anthropic’s Reality Check, The Capital Squeeze, & The New World Order
This week’s issue: for two years the AI build-out was priced as a growthstory. This week it started being priced as a credit story.
A guest post on Tao’s blog asks mathematicians to judge all work as if AI helped, and The New Yorker reports that Tao now describes the era up to 2025 as one of proof scarcity. In October, every AI proof claim we covered arrived citing a Lean check, by our count.
A note on sourcing. This piece relies on Frank Calegari’s guest post on Terence Tao’s blog (1 November 2026) and The New Yorker’s profile of Tao by Rivka Galchen (1 November 2026), both read in full on 2 November, and on FourWeekMBA’s own earlier reporting. The count of October AI proof claims citing a Lean check is ours, from our coverage; we did not inspect any Lean files.
Gennaro Cuofano is a CRO and tech executive who has worked in AI since late 2015, starting with bringing NLP, natural-language generation, voice and chatbot products to the marketing industry. His background is in law and finance: he holds a Master's degree in Law and an International MBA with an emphasis on corporate finance (LUISS Business School and the University of San Diego, 2012), and worked as a financial analyst at a real-estate investment firm in San Diego and as an assistant controller. His work focuses on business model strategy, business engineering and, more broadly, structural analysis: how companies make money, read from their own filings. He created FourWeekMBA and leads research there and at The Business Engineer, his newsletter on AI and business strategy, with over 95,000 subscribers and more than 1,000 published analyses. His writing has also appeared on Entrepreneur (for example, How to Digitize a Family Business in Europe) and HackerNoon. Find him on LinkedIn and Substack. See how we source.
