What should a university protect when a machine can do the homework? A committee at MIT spent five months on that question. It started from its own premise that generative AI, meaning tools such as chatbots that write text on request, can now complete almost any written undergraduate assignment. Its report, dated August 13, 2026, does not recommend better ways to catch cheating. It asks the university to rebuild teaching around what students are supposed to learn.
The hosts of Moonshots with Peter Diamandis discussed the report in an episode published on September 19, 2026. Diamandis, founder of XPRIZE and Singularity University, and his fellow MIT alumni Dave Blundin and Alexander Wissner-Gross mostly agreed with the report's diagnosis. They disagreed sharply about what it means for MIT as an institution.
What the committee said
Eric Klopfer and Sam Madden co-chaired the committee, which based its recommendations on consultation rather than a controlled study. Its work included weekly meetings during the spring, listening sessions, a review of practices at peer institutions and surveys. The committee's own survey received 1,632 responses, a 12% response rate. The report therefore records what members of the MIT community said and what the committee concluded from it. It does not measure how much AI use harms students' thinking.
On the show, Diamandis read from the report. It says AI is "upending foundational elements of the MIT educational experience." That includes problem sets, take-home exams, office hours, study groups and the Undergraduate Research Opportunities Program, known as UROP. According to the report, instructors who have to police unauthorized AI use are damaging their relationships with students, and AI-detection software is unreliable. Diamandis read another line from the report: "an underground river of mutual suspicion is no foundation for a healthy classroom."
Diamandis highlighted what he called the phrase everybody was quoting: "cognitive surrender." He described it as the habit of turning to AI at the first sign of difficulty and letting it take over your thinking. The report also recognizes that AI can be useful for tutoring and research. It argues that the most important product of education is not a grade point average or a diploma. It is a student's personal growth and the development of their own imagination, insight and judgment.
The committee's recommendations follow from that argument:
- Courses should be redesigned starting from their learning goals.
- Assessments should test what students have actually learned.
- Students should still work through hard problems themselves, which educators call productive struggle.
- MIT should protect human contact between teachers and students.
- Instructors should state clear AI rules for each course. The report's sample policies range from requiring AI use to banning it.
The report advises against relying on AI detectors. Much student work now mixes human and AI contributions, and false accusations cause harm. To put the changes into practice, it proposes local AI leads, fellows, support teams and a pilot fund.
"An arms race with your own students"
Blundin, founder and general partner of Link Ventures, said he had not intended to read the report cover to cover but ended up reading every word. "I've never been more proud of MIT," he said, calling it a roadmap for all of education. In his view, a teacher who fights AI in the classroom enters an arms race that wastes everyone's time, the teacher's and the students'. The alternative, he said, is to budget for a large job that never ends. That means rewriting curricula that have gone untouched for decades, then continuing to revise them so students can use AI and still challenge themselves.
His strongest example was UROP, the program that places undergraduates in faculty research labs. Diamandis added that UROP was the reason he went to MIT in the first place. According to Blundin, the report says UROP has drifted into cheap labor for graduate students and professors, though it was meant to be part of students' education. He said the fact that AI can now do some of that work does not change the goal: undergraduates should experience real research and learn from it. "You should not just replace your UROPs with an AI agent," he said.
"MIT should be careful what it wishes for"
Wissner-Gross, a computer scientist and founder of Reified, disagreed. He said MIT "should be careful what it wishes for." He mentioned a mini-manifesto he has written but not circulated publicly, called "The Endless Initiative." It is about MIT's habit of launching entrepreneurship-related initiatives roughly every two years that usually fall flat.
He argued that carrying out the report's vision fully would push MIT toward becoming essentially a for-profit institution. MIT's motto is mens et manus, "mind and hand." If AI automates the "mind" part and every student can get every problem-set answer for free, he said, education shifts to the "hand" part. That would last only until humanoid robots can do hands-on work too, which he expects within a few years. He pictured students launching startups with large economic effects from their dorm rooms. In that world, he said, MIT would start to look like a venture capital fund plus "a bit of adult babysitting." He called his endpoint "my fever dream": MIT and the universities that follow it would become for-profit public benefit corporations listed on the New York Stock Exchange or Nasdaq. That is his own forecast. The report does not propose it.
Labs, robots and the MIT.nano story
Blundin had two answers to that, which he said he had discussed at length with MIT's chairman. The first was physical research. He said AI progress in biotech, nanotechnology, materials and new physics requires a large amount of lab space to generate data that feeds back into AI systems. MIT has a great deal of that space, and in his view the undergraduates working in those labs are hugely valuable to society.
Diamandis raised a challenge. He said he had been on a board call that morning for Lila, which he described as a company building a scientific superintelligence and a million square feet of automated robotic labs driven by AI. Labs like that, he said, will "run circles around" research groups staffed by graduate students unless those students get the tools to direct the AI themselves.
Blundin agreed that MIT could not compete head-on and compared the situation to chipmaking. MIT once built some of the best small experimental chips in the world. Then chipmaking became an industry, and Intel put far more money into it than MIT could ever afford. Research at MIT moved into adjacent fields such as gallium arsenide and later photonics. His advice was the same now: don't compete with Lila, but move into areas it isn't working on, such as materials, nanotechnology or carbon nanotubes.
Wissner-Gross said he agreed with both of them and told what he said was a previously undisclosed backstory. When he was an MIT undergraduate, Business Week ran a story about his nanotech research. Rafael Reif, then assistant head of MIT's electrical engineering and computer science department, called him into his office and asked what MIT could do to be more competitive in nanotech and nanofabrication. Wissner-Gross said he told Reif that MIT needed its own nanofabrication facility. Reif later became president of MIT, and MIT.nano became one of his signature initiatives.
Wissner-Gross says he walks past the building every day and sees relatively little use of its clean rooms. He put utilization at "maybe like 20" and called the building "too clean for its own good." He would like it used at close to 100%. He thinks the best route is for MIT to act more like a Berkshire Hathaway-style for-profit holding company whose assets, including its nanofabrication facilities, could be opened more to outside users. The utilization figure is his informal estimate, and the holding-company idea is his proposal, not an MIT plan.
Purpose first, or a whole new model
Diamandis argued that high school and college both have to be reinvented, and that the new programs share one principle: teaching students purpose. He mentioned TKS, a high school program he is involved in, and said students should learn AI and entrepreneurial skills and start companies. "Your thesis in high school and your thesis in college should be the company you built," he said. That is his proposal, not one of the committee's recommendations.
Salim Ismail, founder of Open ExO, went further. He said most universities began as divinity schools and turned into job-training institutions during the Industrial Revolution. Wissner-Gross interjected that this was true of "all of them other than the ones that didn't," and Ismail conceded that some, including MIT and Waterloo, did not start that way. Ismail argued that education has worked on the supply side: students learn a skill such as medicine or accounting, then look for demand. He said the half-life of a skill has fallen from about 30 years to about three, and no one knows what jobs will look like in five years. He wants schooling at every level to start from the demand side: get excited about a problem, then work on it.
Ismail called the report "one ray of light in the gloom of higher education" and said it points in the right direction. He doubts, though, that established universities can make such a big change. He ranked academia's institutional "immune system," its resistance to change, as the second-worst, behind religion. He argued that the change will come from new models built at the edges.
A neutral place to judge AI
Blundin also gave a second reason he thinks universities remain valuable. AI companies are often suspected of mixed motives, he said, and he pointed to earlier segments of the episode about companies preparing to go public. In his view, that creates a need for a trusted "source of truth" on AI. He thinks MIT should fill that role by staying politically neutral and focused on the long-term good of humanity. He criticized the university for having "got into politics" for a while. He argued that because students of every nationality and religion come to MIT, work on AI ethics is better done there than in a for-profit company. If MIT builds on both its physical labs and that ethical role, he said, it will have a "very, very important role going forward."