100+ reactions to 100+ solutions
100+ reactions to 100+ solutions
Source: Proofs and Prompts
Date: October 8, 2026 (Updated October 9, 2026, 10:15 pm CEST)
Note from the Hosts: Following the OpenAI announcement on October 6, many of us are struggling to process the implications. We believe it is vital to capture a "gut-feeling" snapshot of the community's current state. This collection will remain open for one more week. To emphasize our collective experience, contributors are listed by name only—omitting professional titles or affiliations.
If you wish to contribute your own perspective, please contact proofsandprompts@gmail.com.
The Impact Landscape
The reactions range from awe at the technical achievements to existential dread regarding the future of the profession.
Community Perspectives
| Contributor | Primary Sentiment | Key Focus |
|---|---|---|
| Tim Santens | Astonishment | The necessity of a research transformation. |
| Roman Sauer | Analytical | The AI's use of specific human "tricks." |
| Xiaolei Wu | Shock | The collapse of long-standing conjectures. |
| Nicholas Williams | Melancholy | The shift from solving to understanding. |
| Henry Bradford | Bewilderment | The crisis of mentorship and apprenticeship. |
| Henry Wilton | Critical | The destruction of academic structures. |
Detailed Reflections
Tim Santens
The output from OpenAI is nearly impossible to believe. Specifically, the result regarding Quasi-Riemann felt like something out of science fiction. While one could critique the ethics or behavior of AI labs, the mere existence of these models means that mathematics, as a research field, is facing a gradual change massive transformation.
Roman Sauer
There is a common belief that mathematicians rely on a small set of versatile "tricks." I have one such method: replacing a manifold with a specific measurable foliated object (derived from ergodic theory) and applying it in surprising ways. I used this in a paper with Sabine Braun on macroscopic scalar curvature, which Hannah Alpert later expanded.
Remarkably, this exact approach appears in OpenAI's creative solutions for problem 207 and problem 335—two problems that are entirely unrelated. We cannot simply protect our ego by narrowing the definition of "human" mathematics; we must radically broaden our understanding of what a mathematician actually does.
Xiaolei Wu
I discovered the news on the final day of a national holiday in China, while returning from a family trip to the coast. I woke up to a flood of PDFs. I had previously attempted these problems using ChatGPT, but I likely lacked the necessary tokens or the latest model version.
The implications were staggering:
- The Eilenberg–Ganea conjecture is false.
- The ambient simple group could be of type .
- The Borel conjecture fails in dimension 4.
- The coarse Baum-Connes conjecture also fails.
The rest of my day felt endless. While some on social media tried to downplay the importance of the solved problems in their respective fields, the reality is hard to ignore.
Nicholas Williams
While we have gained a wealth of new facts, the process of digesting these papers will be slow. LLMs have fundamentally altered the landscape, but the ultimate objective must remain human comprehension.
Mathematicians do more than just clear a backlog of open problems; we determine which conjectures are actually worth pursuing. However, it feels like the end of an era where the ability to solve a hard problem was the primary marker of a mathematician. For those of us who love the act of solving, this is a sad realization.
Henry Bradford
I am currently feeling a mix of disorientation and bewilderment. The speed of these developments has surpassed our ability to formulate a response. My deepest concern is for the next generation.
def career_path(student):
effort = "years of refining craft"
incentive = "ability to contribute new knowledge"
if incentive == "replaced by AI":
return "crisis of motivation"
return "meaningful contribution"
I worry that the incentives for bright young people to invest the necessary years of effort into the craft may vanish. I hope for a future where AI enriches our understanding of the mathematical universe rather than undermining it, but we need a realistic new pathway for young scholars to establish themselves.
Henry Wilton
I have a deep personal interest in several problems on the list, including:
- The Cannon conjecture
- The non-residually-finite hyperbolic group
- The Boone—Higman conjecture
- others
A quick look suggests the quality of the write-ups is inconsistent. Verifying and digesting these proofs will take months. If the goal was to dismantle the mathematical community, this "data dump" is the perfect weapon. Open problems are a shared resource that provide a yardstick for progress. This isn't just a shock; it's a tsunami. OpenAI seems willing to erase our community's norms—such as the necessity of naming authors—to serve their own financial goals.
Contributor Avatars
