Unfixed Newsletter — September 17–30, 2026
Editors’ Note: A couple high profile cases of AI use by faculty and administrators have foregrounded a different conversation about what is appropriate. Honestly, we are overdue for a glance in the mirror at our own AI usage as a profession. Outside of this we have a breakout consumer AI agent, a possible infrastructure slow-down and a viral interview. In a slower news cycle any of these items would have been a lead theme, but this has been another intense two weeks.
The Stories
1) Dartmouth’s provost becomes a test case for AI disclosure at the top of the university
Dartmouth has opened an independent review of Provost Santiago Schnell’s use of AI in academic publications and public writing. The review follows reporting by The Dartmouth that found extensive signs of AI involvement in Schnell’s recent work. Schnell has acknowledged using AI for editing, organization and language refinement and now says he should have disclosed that use more consistently.
The case arrives alongside a broader administrative trend. Inside Higher Ed’s new survey of 376 chief academic officers found that roughly seven in 10 provosts use AI in their own work at least weekly. Seventy-one percent use it to summarize documents and reports, 66 percent for presentations and meetings, and 65 percent to draft communications to faculty, staff or students. Only about one in 10 said their institution has a centralized AI strategy.
Why this matters: This is a high profile case of what we think will be a real trend in the coming months. There are unresolved questions about when AI is appropriate for students, but now we have to answer some of those same questions for ourselves. Some of these cases will be about accountability, some of them will be little more than witch hunts, and it is to be determined how much people in and outside the academy actually care. Nik and I are working on a piece about this exact issue we hope to share with a broader audience soon.
Links:
https://president.dartmouth.edu/news/2026/09/review-and-community-conversation-ai
https://www.thedartmouth.com/article/2026/09/schnell-ai-writing
https://www.insidehighered.com/news/governance/executive-leadership/2026/09/23/ai-use-funding-cuts-how-provosts-navigate-2026
2) One professor produced 200 papers in nine months
University of Chicago statistician Nicholas Polson authored or co-authored more than 200 papers this year, plus as many as 14 books, while using AI as part of his research process. The output drew enough scrutiny that SSRN removed 257 works associated with Polson and froze the accounts used to upload them.
Polson has not been accused of fraud and says AI allows productive researchers to move faster. The scale is what makes this case difficult to dismiss as another disclosure dispute. Journals and reviewers were already dealing with rising submission volumes. AI may now be capable of changing the economics of academic production itself, including how much scholarship can be generated and how much human attention remains available to evaluate it.
Why this matters: There have always been discrepancies in publication records as a result of different resources. A professor with a legion of graduate students and summer support will almost always produce more than someone from a state school with a heavy teaching load and no resources. However, this is an order of magnitude larger than anything we have seen and we need to accelerate the pace of conversation about this on college campuses. As a side note, DJ Hopkins and I have a piece out in Academe Blog addressing some of these issues.
3) Meta’s Muse looks like the first breakout consumer agent
Meta launched Muse a consumer friendly AI agent earlier this month, but the story accelerated during this issue window. Muse quickly reached the top of U.S. app-store rankings. Muse is built to access other apps and services rather than wait inside a chatbot window: email, calendars, shopping, travel, and other everyday tasks. A user can grant Muse access and the AI agent will carry out tasks for you. Amazon already moved to block the agent from shopping on its site. Amazon’s response shows the next agent problem already arriving. Users may give agents permission to act for them while the systems on the other side refuse to let those agents in. Either way, AI agentsI have a clear use proposition: get things done for you so you can do something more important.
Why this matters: We have long theorized that AI agents are the next big thing, with ramifications for higher ed. Most AI agents rely on more complex software such as OpenAI’s Codex or Anthropic’s Claude Cowork, or insanely complex do-it-yourself OpenClaw set ups. Muse, on the other hand, is one of the first plug and play free apps on your phone. While we don’t think Muse can complete course work for students (yet) it certainly is the first AI agent to breakthrough into the mainstream. Students, staff, and faculty can now have Muse triage email, calendar, documents, and other apps on your phone or Mac. Maybe we’ll suddenly either get dozens of more meeting requests or maybe our agents will all just figure it out themselves. Kidding aside, Nik tested Muse over the weekend and ran into the biggest red line: does he trust Meta with all his personal data across all his apps and services? His answer was a big NO. Muse will be going back in the drawer. But it sure looks like agents are about to have their moment, for better or worse.
Links:
https://about.fb.com/news/2026/09/introducing-muse-personal-ai-agent/
https://www.techradar.com/ai-platforms-assistants/i-gave-muse-access-to-my-personal-data-and-now-i-know-why-personal-agents-are-probably-going-to-change-the-world
https://about.fb.com/news/2026/09/the-biggest-news-from-connect-2026/
https://www.axios.com/2026/09/21/amazon-meta-muse-ai-agentic-shopping
4) The AI slowdown conversation reaches the money
The calls to slow frontier AI development that dominated the previous news cycle have begun colliding with a second question: whether the extraordinary financial buildout behind AI is sustainable.
SoftBank-backed data-center developer SB Energy delayed an IPO that had targeted a valuation around $50 billion after struggling to attract enough investor demand. Reporting this week also pointed to pressure in AI-linked debt and growing scrutiny of the hundreds of billions being committed to compute infrastructure.
Why this matters: We steer away from speculation about bubbles and financial markets because we really don’t have expertise to inform commentary. That said, financial markets, anti-AI political winds, and calls for lab regulation are all pointing toward some sort of slow down. For us in higher ed this matters because we may be in contractual relationships with these labs and this might start to impact the availability of services.
Links:
https://www.ft.com/content/4a80f076-d1bb-4492-a854-0833d981d5f4
https://www.marketwatch.com/story/the-hottest-part-of-the-ai-trade-could-be-turning-into-its-biggest-weakness-099f2c09
5) Jensen Huang makes the maximalist case for AI to Ezra Klein
This one is less a traditional news event than a piece of the AI debate that has traveled unusually far, including through academic circles.
On the September 23 episode of The Ezra Klein Show, Nvidia CEO Jensen Huang laid out one of the clearest versions of the optimistic case for rapid AI development. He rejected prominent estimates of catastrophic AI risk as unsupported by science, argued against broad new regulation, and pushed back on predictions that AI will destroy more jobs than it creates.
Klein presses him on a recurring tension in that position: even if new jobs eventually emerge, transitions can still be brutal for particular occupations, communities and generations of workers. That exchange is one reason the interview has been circulating beyond technology audiences. It gets directly into questions colleges are already struggling with about expertise, entry-level work, professional preparation and what students should be learning if AI can increasingly perform parts of skilled jobs.
Why this matters: This interview was incredibly wide ranging and covered nearly every aspect of this AI moment from infrastructure build, to safety, to the future of the economy. There were a few exchanges that got a lot of attention from academics. About 20 minutes in there is a fascinating conversation about the future of engineering where Huang reflects on the difference between his talents as a new employee in comparison to new college graduates in whom he identifies “systems thinking” as a strength. The thing many people latched onto was his claim that arithmetic and the memorization of data (even his own address) may not matter in the future. This along with speculation about the future of the economy really set academic social media on fire. The interview is absolutely worth your time to listen to.
Links:
https://www.nytimes.com/2026/09/23/opinion/ezra-klein-podcast-jensen-huang.html
https://www.axios.com/2026/09/23/nvidia-jensen-huang-ai-doom-predictions
From Our Work
You Already Work at Two Universities
September 21, 2026
Zach Justus and D.J. Hopkins write about the widening divide between faculty who are incorporating AI deeply into their work and those who are not, and what that divide could eventually do to workload and productivity expectations.
Ep. 40: Did the Student Learn? Rethinking Education in the Age of AI with John Katzman
September 21, 2026
Nik and Zach talk with Princeton Review founder and Noodle CEO John Katzman about AI tutoring, evidence of learning, online education and whether authorship remains the right question for evaluating student work.