Shiwen Dialogue Issue 31 | Mathematicians are crying, the foundation shelled out $15 million: Emotional costs are officially put on the books for the first time

Opening Dialogue: Machines Demolished Mathematicians’ Towers

Shiwen: This week’s most unusual news in the tech world isn’t about a new model topping the rankings, but about a group of mathematicians publicly saying they are crying. Let’s start from the beginning?

Yongliang: Okay. The timeline is short. At the beginning of October, it was reported that OpenAI released hundreds of mathematical discoveries, many proofs spanning number theory, theoretical computer science, and logic, some of which had stumped the world’s best mathematicians for decades or even centuries. On October 15th, the Simons Foundation announced a partnership with Anthropic and OpenAI, committing a total of $15 million in funding and resources to help the mathematics community cope with the rapid rise of AI. On one side was industry-wide sorrow, on the other a compensation, less than two weeks apart.

Shiwen: Why is it unusual?

Yongliang: Because we’re used to two narratives: either “AI replaces certain professions” or “AI is a tool.” This time, the impact was felt by a group who see themselves as people of intellect, and the companies that affected them chose to pay for workshops and set standards, marking the first time “emotional cost” was recorded in accounts. I’ve led technical teams for seventeen years, managing seventy people at peak times, and I know what such records mean. My personal experiences will be highlighted here.

Q1: What Are Mathematicians Really Crying About?

Shiwen: Let’s clarify the sorrow itself first. What are the key quotes?

Yongliang: Two sentences are most significant. Hugo Duminil-Copin, a Fields Medalist in 2022, said in an interview, “I cried on Tuesday night, waking my daughter. They destroyed my research field.” Cornell mathematician Steven Strogatz put it more complexly, “People use phrases like ‘they mining our discipline outdoors’ or ‘they detonating this landscape.’ It feels like being lost. Some days I’m dizzy, some days I’m excited. There’s also immense sorrow.” Note that Strogatz didn’t just talk about sorrow; he combined excitement with dizziness, a detail we’ll discuss later.

Shiwen: A Fields Medalist, the highest honor in the industry, says they were “destroyed.” This is not ordinary discomfort.

Yongliang: Right, what was destroyed wasn’t just a job, it was a tower. The field that mathematicians like Duminil-Copin have built over decades was flattened by machines in just a few months. Derek Thompson, who wrote “I Know Why the Mathematicians Weep,” made a judgment I agree with: rapid progress in intelligent technology is destined to clash violently with modern sentiments that value human intelligence, often equating it with purpose, dignity, and self-worth. In plain terms: When your professional identity is built on “I’m the smartest group,” machine intelligence isn’t competition, it’s an identity crisis. And identity crises don’t have training courses.

Shiwen: Have you seen similar scenarios when you led teams?

Yongliang: I’ve seen it before; this is my experience, not statistics. The Java ecosystem has gone through several major transformations, such as frameworks shifting from SSH to Spring Boot, and deployment moving from physical machines to containers. There were always a group of engineers in their forties with the highest code quality, but their skills suddenly became less valuable. They say they can’t learn anymore, but actually, it’s their identity that gets stuck: What I’m proud of can be done by machines and new frameworks in ten minutes—so who am I? The skill and dignity of engineers are completely equivalent to the glory of mathematicians in this regard.

Q2: What was bought with $15 million in compensation?

Shiwen: Let’s look at the money. What exactly is this $15 million? Yongliang: According to the Simons Foundation’s official press release, the three parties agreed to provide $15 million in funds and resources, without disclosing individual shares or the ratio of cash and computing power. The funds are managed and distributed by the foundation, including support for public proposals, workshops, and initiatives, helping researchers evaluate recent AI-assisted results, learn new tools, exchange ideas, and establish collective standards for AI use. Anthropic and OpenAI also provide AI tools and computing power to researchers, covering areas beyond mathematics to basic science. Shiwen: Why is the Simons Foundation leading this? Yongliang: It’s not an outsider. The foundation was founded in 1994 by Jim and Marilyn Simons. Jim Simons established Renaissance Technologies hedge fund. According to Wikipedia, the foundation’s assets exceeded $5 billion in 2022. More importantly, its history: the official statement describes it as an early supporter of scientific machine learning, having funded the development of Lean proof assistants. It understands the mathematical community and has long bet on AI-assisted mathematics. Having it lead this funding means that the mathematical community is least resistant to this. Shiwen: What did the chairman say about the purpose of this money? Yongliang: David Spergel’s two sentences are worth reading carefully. The first: “As AI rapidly reshapes the research landscape, we want to help the mathematical community unite and find the best path forward. It’s time to view mathematics as a whole: not just a collection of proofs and theorems, but a shared understanding driven by human curiosity.” The second is more direct: “How these tools are evaluated, adopted, and used should be driven by the community, not any single company or institution.” In other words, OpenAI and Anthropic provide the money, but the rules should be determined by mathematicians themselves. This is a design to bring those affected back to the negotiating table. Shiwen: From your management experience, what is the real purpose of this money? Yongliang: The money itself doesn’t solve the identity problem. $15 million is not much for a foundation with $5 billion in assets. The real signal is that it acknowledges: organizations are willing to officially admit that “professionals have emotional costs and deserve proper compensation.” Over the years of management, I’ve learned this from experience: when teams are impacted by technological changes, the hardest thing is not training budgets, but the lack of recognition for their difficulties. Recognition itself is intervention.

Q3: Skepticism in the comments section: Is this compensation or a way to whitewash issues?

Shiwen: This news has received over 200 likes on Hacker News. Do people agree with this in the comments?

Yongliang: I’m not very convinced. Let me relay the main points. Some say that OpenAI and Anthropic are being criticized for turning mathematical discoveries into PR wins for their models. Now, each is spending around $5 million (this is an estimate from the comments section, with no official figures disclosed) to clean up their reputation. Some of this money is in the form of tokens. The foundation is essentially giving the laboratories a “decent cover” for their mistakes, possibly the fastest-approached funding proposal ever. Some are more direct in their emotions: one of the foundations responsible for the AI chaos has even bribed mathematicians to speed up AI development. Others are worried about distribution: the money likely goes to already well-funded top mathematicians, while early-stage students get the chance, like donating to the Node.js foundation instead of those who really need support.

Shiwen: Is there any doubt about the amount itself?

Yongliang: Yes, and in two directions. One argument is that $15 million isn’t much; half would go to university administration, leaving little for research groups, possibly only enough to fund less than ten teams. Another point is that this amount is “change” for these two companies, and using a large number as a headline is suspicious. The most heartbreaking question is: if the tokens are given to mathematicians, will it just encourage more models to be used, which is the problem itself? Practical experts suggest turning off the data authorization for “help improving models” in ChatGPT and Claude settings, although published works can still be scraped.

Shiwen: How do you rule on these doubts?

Yongliang: Motivational doubts and value of the mechanism can be separated. Motivationally, the companies definitely have self-interest, which doesn’t need to be addressed. Mechanism-wise, workshops, public proposals, and community standards are real benefits for researchers. The official document states that “discussion is driven by the community, not a single company,” which is a commitment to accountability. My judgment: look at the deliverables. Whether proposals were solicited, how many workshops were held, and whether AI usage standards were documented will be clear within three months.

Q4: The Other Side: Sadness Is Not the Industry Conclusion

Shiwen: Has anyone said that mathematicians cry too early?

Yongliang: There are some technical boundary discussions, and I think their quality is not low. One commentator said that AI essentially performs fuzzy testing, searching for valid inputs in the solution space. It doesn’t understand what’s happening, and without understanding, it can’t create the next generation of mathematical abstractions. Another said that large models are just暴力 solving within existing knowledge, which will eventually run out. Then the only way is to train them with newly discovered human knowledge, but finding new knowledge may be exponentially more difficult. A mathematician who claimed to be honest said: I’m sad too, but in a sense, my sadness has already been included in the price. Changes are indeed fast. Someone recalled that mathematician KConrad was saying in 2024 that AI was useless for research. By May this year, AI had overturned the unit distance conjecture in just five months, surpassing humans in almost every aspect.

Shiwen: Both sides make sense. Where is your balance?

Yongliang: Back to Strogatz’s own words: Some days I feel dizzy, some days I’m overjoyed. No one involved can settle on a single emotion, let alone outsiders. The theory of technical boundaries answers the question “Can AI create new mathematics”? There’s no answer to that yet. Thompson’s rhetorical question is honest: What if we should view mathematics primarily as a quest, not just a dialogue equally important to feeling and discovery? Strogatz said that Archimedes wrote two thousand years ago that there’s no shortage of newcomers to contribute to the theorems we’ve established. Mathematics is a dialogue between generations, not just results. Whether the dialogue is interrupted by machines brings excitement or sorrow, depending on where you place yourself in it. This is an opinion, not a conclusion.

Shiwen: Thompson also has a sentence that’s frequently quoted in the comments section.

Yongliang: “Today, mathematicians are crying because machines first destroyed their towers. But tomorrow, those crying might be physicists, economists, illustrators, or writers.” He added his own self-positioning: I’m not the kind of AI theorist who believes that hundreds of millions of jobs will be eliminated within five years, but I’m a realist. This is also the tone I set for this issue: No selling the apocalypse, nor pretending nothing is wrong.

Q5: For professionals and managers affected by this

Shiwen: How does this apply to specific individuals?

Yongliang: Three sentences, all based on experience, not statistics. To professionals: Sadness is rational; it’s a清算 of sunk costs and the beginning of revaluation. Don’t rush to “get out,” but set a deadline for mourning your skills. After that deadline, assess what remains beneath the damage: the ability to ask good questions, the taste for judging outcomes, and cross-disciplinary intuition. Machines learn to solve problems first, then to ask questions. To managers: When team core skills are normalized, you need to manage not just skill重置, but also identity reset. Training budgets address the former, acknowledging emotions address the latter. Both need to be prioritized; don’t approve only one. To everyone: Notice if anyone in your industry is starting to establish “AI usage standards.” The $15 million initiative by foundations is formalizing the recording of professionals’ emotional costs. Other industries will likely follow, and early entry gives the most valuable control over rules.

Shiwen: The last question: What is this most easily misconstrued as?

Yongliang: It’s most easily misconstrued as “mathematics is over.” No. This column reported last week on the verification issues of AI mathematical achievements. Here’s a bit more: Hundreds of achievements are still being audited, and the community’s first reaction is to establish evaluation standards rather than surrender. This itself shows that the industry still has strength. The real signal isn’t mathematicians crying; it’s foundations funding mathematicians to decide how to use these tools. Crying is emotion, funding standards is action, and the latter represents the direction of the industry.

Conclusion

Machines first destroyed mathematicians’ towers, then those who destroyed the towers funded it, asking those who were damaged to sit down and set rules. This issue’s focus isn’t “Will AI replace mathematicians”? Whoever answers that question now is guessing. The focus is on that $15 million initiative: for the first time, professionals’ emotional costs are officially recorded in the ledger. Sadness is rational; organizations that acknowledge sadness are smart. When the next crying industry appears, see if anyone follows suit—that will tell if this is an isolated case or a paradigm shift.

Depth: Three topics to follow

1. The delivery of the proposal solicitation. Whether public proposals were accepted, how many workshops were held, and whether collective standards for AI usage were established are hard indicators of the sincerity of this funding, which will be visible within three months.
2. The hierarchical flow of funds. Whether the concern in the comments section about “top mathematicians receiving funding and students taking over” becomes reality depends on the distribution of institutions and ranks among the first batch of recipients.
3. Follow-up from other industries. Which industry—physics, economics, illustration, or writing—will first experience its own “Simons Moment” will mark the transition of emotional cost accounting from an isolated case to a paradigm.

Information Sources

• Simons Foundation official press release (2026-10-15, direct source): A total of $15 million in funds and resources from three parties, public proposal solicitation, workshops, and collective standards, two quotes from Spergel, the foundation’s establishment in 1994, funding for Lean proof assistant, extending to basic science.
• Derek Thompson’s “I Know Why the Mathematicians Weep” (2026-10-15, HN 325▲): OpenAI released hundreds of mathematical discoveries in October, quotes from Duminil-Copin and Strogatz (originally from a Wired interview, cited by Thompson’s article, marked as “interviewed”) and an excerpt from Archimedes, Thompson’s judgment sentence.
• HN story 50104654 (Simons program, 215▲): All doubts in the comments section were repeated without naming specific individuals; “each contributing approximately $5 million” was labeled as an estimate from the comments section.
• HN story 50103630 (Thompson’s article, 325▲): Discussions on technical boundaries, bananaflag, and QuesnayJr’s views, summarized and cited.
• The foundation’s assets exceed $5 billion: A commenter on HN cited Wikipedia, which is non-official content, with the text marked accordingly.
• Guest opinions are based on personal experience and estimates, not statistical conclusions.