Editors’ note: Our lead story this week is a huge investment by Stanford in the sort of course/curriculum revision work we have been engaged in for the past few years. Part of why we are leading with this is because it has not gotten a ton of attention in other spaces. There is a lot going on right now, but this story is hugely important as an elite institution recognizes that AI is changing their world, not just ours. We hope this adds volume to a much-needed national conversation about what higher ed looks like now that AI has arrived and continues to accelerate.
The Stories
1) Stanford Plans to Invest $10 Million to Redesign Majors Around AI
Stanford announced a $10 million initiative to help academic departments redesign undergraduate majors in response to AI. Beginning in winter 2027, departments can apply for grants of up to $750,000 to reconsider course sequences, learning outcomes, and assessments. The initiative asks departments to determine which skills students should develop using AI and which they should demonstrate independently. It also allows for approaches that deliberately exclude AI, including in-class activities and oral exams.
Why this matters: This is straightforward and also significant. Stanford is the latest elite institution to recognize wholesale changes to curriculum are necessary. First, the funding is significant and a recognition of the scale of changes necessary. Second, sometimes moves at elite universities compel other universities to follow so we should not view this as a “well that is Stanford” story. Finally, combined with the MIT study from earlier this year there is a strong signal that even flagship campuses are recognizing the urgency of the moment. We have been outspoken about not over-indexing on elite institutions, but in this case we hope leadership at other campuses take notice.
Link: https://news.stanford.edu/stories/2026/11/provost-initiative-college-majors-ai
2) Anthropic Releases New Claude Models as Researchers Find AI Accelerating Its Development
Anthropic released Claude Fable 5.2 and Mythos 5.2 on November 9, reporting improvements in coding, reasoning, and scientific research. The company says the models helped solve two previously open physics problems and design three working electronic devices. Days earlier, independent research organization METR reported that Anthropic's model capability benchmarks had improved at approximately 1.9 times their previous rate since October 2025. METR attributed much of the acceleration to AI-assisted research but acknowledged substantial uncertainty, compounded by Anthropic's redaction of supporting evidence.
Why this matters: As with many AI breakthroughs in the past few months there is reason to be cautious and wait for some of these larger claims to be verified. The movement into engineering is novel and part of why we included this. Previous model progress has largely been oriented around theoretical math, physics, and computer science, but this is a potential bridge to the physical world with trickle down implications for society and education. This model release is related to how fast the labs are accelerating and this report from METR. If you are unfamiliar METR is an extremely trusted organization and they are often who the labs turn to when things go wrong as they have the trust of the public and in-house experts. They do have close ties to Anthropic and some commentators think that puts their statements on the company in question. This does not rise to the recursive-self-improvement threshold we keep hearing about, but it does help to explain the increased pace of change from the labs with every week bringing new models for coding, text, images, and video. We think this connects to the Stanford story because our adaptations to AI need to be flexible to account for this pace of change. Also, we need to redevelop systems–not just policies, to make rapid adaptation possible.
Links:
https://www.anthropic.com/claude-fable-and-mythos-5-2
https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleration-assessment/
3) Nvidia Investigates CEO Jensen Huang Over Alleged Chinese Espionage Operation
Nvidia's board announced an independent investigation on November 8 after reports that CEO Jensen Huang had been targeted by Chinese intelligence. Huang acknowledged an extramarital affair with Alice Huang, a former executive at a Singapore-based chip company whom U.S. intelligence officials reportedly suspect of being recruited by China. China denies the allegations. Nvidia says it has no reason to believe company secrets or national security were compromised. Independent directors and outside counsel will conduct the investigation.
Why this matters: This is a reminder that our legitimate concerns about education and AI are just part of this puzzle. We often feel like the labs are releasing things just to make our lives more challenging, but this story brings the scale of this challenge into focus. There are implications for relationships between super-powers, questions about how much influence the CEO of the most valuable company of the world should have, and even a disruption to global markets. This is a callback to the great blog piece Nik wrote a while back on why AI is not Edtech. When we call Instructure because Canvas was hacked or have product questions about Turn-it-in, we are their primary concern. When we want features updated in ChatGPT we have to realize they are also fielding calls from world leaders, global financial institutions, and intelligence agents from other countries. Our access and influence is not the same as we are used to.
Links:
https://nvidianews.nvidia.com/news/statement-from-nvidia-board-of-directors
4) Harvard Study Finds AI-Generated Academic Reviews Outperform Human Reviewers
Harvard researchers Gary King and Steven Worthington compared 134 expert-written reviews of 124 social science manuscripts with reviews generated by Claude. An independent AI evaluator preferred the AI-generated review in every comparison, while three former journal editors preferred AI reviews in 87 percent of their assessments involving ten review pairs. The research examined developmental feedback provided before formal journal submission, rather than the complete journal peer-review process. The paper has not itself undergone peer review.
Why this matters: Peer review has long been the gold standard for scholarship because expert perspectives are important in vetting new knowledge. Peer review has always had problems including gatekeeping by editors, long wait times, and politicking. This system had already started buckling in the past year due to an influx of AI written/assisted manuscripts. Now we have to ask questions about who is writing scholarship, who should be reviewing, and what the place of expertise is in the process. A note of caution, this study itself has yet to be peer reviewed so we should consider this advance notice of a possible direction rather than a certainty.
Links:
https://gking.harvard.edu/peerless/
5) Voters Overwhelmingly Support Restrictions on Data Center Development
Voters favored limits on data center development in 40 of 41 local ballot contests on November 3, according to Ballotpedia. Twenty-seven measures restricting development or expressing opposition passed, including five full bans, twelve partial bans, and three temporary moratoriums. Voters also rejected thirteen measures that would have supported data center development. The measures addressed zoning, electricity and water consumption, local approval requirements, and development restrictions. Some results were nonbinding expressions of voter preference rather than enforceable regulations.
Why this matters: Surveys and opinion polls are one thing, elections are something else. This is the clearest signal we have seen that public opinion is sharply against data centers and specifically AI data centers. Campuses and systems (including ours) have entered into partnerships with AI labs and technology companies and we are very on-the-record about our views on it. The new issue we have to navigate is explaining this to the public and to potential students in a way that isn’t alienating. All indications are this will continue to be a salient political and social issue for the years to come so the time to reconcile this is now.
From our work
Using AI While Writing Without Having it Write for You
Nik Janos
Can AI help you write without doing the writing? We share a practical workflow that uses AI to turn recorded conversations into outlines while keeping the ideas, words, and authorship human.
https://www.meltsintoair.org/chatgpt/using-ai-while-writing-without-having-it-write-for-you
Unfixed Podcast, Episode 44: The Coming AI Detection Wars
Nik and Zach examine AI detection controversies involving a novelist, an international literary prize, and a university provost, including questions of authorship, disclosure, and trust.
https://www.meltsintoair.org/unfixedpodcast/ep-44-the-coming-ai-detection-wars