Unfixed Newsletter — October 1–14, 2026
Editors Note: The publication of hundreds of math papers by OpenAI dominated headlines in this cycle following their promotion of a possible solution to the longstanding Navier Stokes conjecture which we covered and wrote about previously. Combined with stories about humanism and the politics of representation we are seeing a new theme–the integration of disciplinary areas into this conversation we have not seen before. We hope institutions see this as an opportunity to lean on faculty who may have felt marginalized in other parts of the AI and higher ed conversation.
The Stories
1) OpenAI releases 700 AI-generated math papers, and some mathematicians are furious
On October 6, OpenAI released more than 700 AI-generated mathematical manuscripts addressing hundreds of open problems. The announcement drew sharply divided reactions. Some mathematicians welcomed the possibility of significant new results, while others criticized the volume of largely unverified material and the burden of evaluating machine-generated research.
Reporting in Nature documented concerns about the quality of the manuscripts, the difficulty of checking their claims, and whether releasing such a large collection constitutes a meaningful scientific contribution.
Why this matters: There are layers to this. Mathematics researchers and professors may have spent years or careers trying to make progress on some of these problems only to see that work disrupted overnight, often in ways that are difficult to verify. This also portends an uncertain future for many fields. In response to this flood of research Ethan Mollick tweeted “I am seeing rapid increases in the ability of AI to do novel work in my field of economic sociology, with nearly autonomous research getting to top journal level.” We are fooling ourselves if we think uneven advancement isn’t coming to other fields. In the academy we may need to shift resources from research to verification if our goal is advancing knowledge in our fields. We also may need to reconsider the function of our own research and ask whether we are trying to advance our fields, or using research as verification of our own expertise, or perhaps something else. It is a new world.
Links:
https://openai.com/index/sharing-ai-progress-in-mathematics/
https://www.nature.com/articles/d41586-026-03196-8
2) Ezra Klein continues to bring deep AI conversations to a broad audience.
The last newsletter featured a Klein interview that had broken out of normal conversation frames; this newsletter has two. The Ezra Klein Show offered competing assessments of AI's trajectory. On September 29th (it broke out during this newsletter window) Bill Gates offered some dire warnings about AI. On October 7, former OpenAI safety leader David Robinson discussed his departure from the company and concerns about its approach to AI safety.
These conversations brought some of the AI safety conversations to different audiences in new forms. We think they are both worth listening to and provide valuable counterpoints to the interview with NVIDIA Jensen Huang’s interview from the last cycle
Why this matters: The AI safety debate implicates everyone. But beyond our status as fellow humans, the AI safety debate matters differently for higher education. Near term safety risks like data security are especially important for institutions beholden to FERPA and conducting research. There is also the question of whether institutions that have contributed to the development of LLMs should have different responsibilities in this conversation. Finally, Gates is a controversial figure especially in higher education because of Epstein associations, outsized influence on education at all levels, to say nothing of his anticompetitive history in business. However, he has some insights in his interview about educational innovation that are worth considering. Gates makes an excellent point about the difficulty in scaling innovation outside faculty who are predisposed to innovation. Even for people working in this space for a while, it gave both of us pause.
Links:
https://podcasts.apple.com/us/podcast/bill-gatess-blunt-warning-on-a-i/id1548604447?i=1000792163003
3) Stanford's AI-photo scandal becomes a student-governance fight
Stanford University acknowledged in September that its Residential & Dining Enterprises division used generative AI to alter a promotional photograph of students, including changing their appearances and racial representation. The university confirmed that the edits violated its policy governing AI-generated depictions of university people and events.
On October 1, Stanford's Undergraduate Senate introduced a motion calling for greater transparency about the university's investigation and student representation on its AI advisory committee. The controversy has prompted broader scrutiny of how Stanford oversees generative AI use in university communications.
Why this matters: This started off as an embarrassing story at the intersection of an elite institution, a controversial technology, and the politics of inclusion. This news cycle saw a migration into shared governance. It opens a big question about whose responsibility it is to set policies about AI. We have seen academic senates and university leadership attempt to assert control, but this is one of the first instances where student government has stepped in. If this is a trend instead of an isolated incident we will see an intersection between this and the growing antipathy towards the technology among young people. This is one to watch for sure.
Links:
https://stanforddaily.com/2026/10/01/ugs-student-privacy-ai-controversy/
https://stanforddaily.com/2026/09/23/university-confirms-ai-edited-photo/
4) Anthropic wants you to be nicer to Claude. It's also consulting religious scholars about AI consciousness.
On October 8, Anthropic updated its usage policy to prohibit sustained, purposeless cruelty toward its AI models. The company says ordinary frustration, criticism, and testing are exempt. Claude can end abusive conversations, while repeated policy violations could result in account restrictions. The announcement quickly attracted widespread attention and ridicule online.
The policy follows a September 29 New York Times investigation revealing that Anthropic has privately convened religious scholars to discuss AI consciousness, moral status, and the ethical character of its models. Participants brought different religious and philosophical perspectives to the discussions and disagreed about whether AI systems could possess genuine subjective experience.
Why this matters:This is another story with academic layers. First, most higher ed institutions include humanities scholars in religion and philosophy. These are some of the folks who may have felt the most marginalized with the breakout of AI disrupting writing. This may be an opportunity for institutions to recenter the expertise of these faculty. Second, if this sticks, higher education institutions will need to add a new wrinkle to existing AI literacy modules and programs concerning model etiquette.
Links:
https://www.anthropic.com/news/2026-usage-policy-update
https://www.nytimes.com/2026/09/29/us/anthropic-claude-morals-ai.html
From Our Work
Ep. 42: What Comes Next for AI in the CSU? with Ed Clark
We speak with CSU Chief Information Officer Ed Clark about extending the system's OpenAI agreement, AI access as an equity issue, cost concerns, possible alternatives including open-source models, and the role of faculty in shaping the CSU's AI strategy.
Ep. 41: The View Without the Climb: Skills in the Age of AI with Ciarán Dunne
We speak with Dublin City University's Ciarán Dunne about developing and assessing transversal skills, including critical thinking, creativity, ethical decision-making, and data literacy, across academic programs.
Stop Expecting AI Companies to Behave Like EdTech Companies — Nik Janos
Nik examines the mismatch between higher education's teaching objectives and the priorities of general-purpose AI companies, including how features designed to reduce work can conflict with assignments intended to develop skills.
We Learned the Wrong Lesson from the AI-Frontier Math Controversy — Zach Justus
Zach examines the controversy surrounding AI-assisted advances in frontier mathematics and argues that disputes over credit and competing AI labs have overshadowed the significance of the scientific work.