This week's theme is guarding the hard parts. As the semester opens, institutions are deciding — in memos, committee reports, and syllabi — which pieces of thinking must remain effortfully human. Chicago's social-sciences Core goes fully analog in the same term its students receive free Claude accounts; MIT answers with redesign rather than retreat; the New York Times asks what is lost when students stop writing; and a high-school English teacher offers the most usable classroom rules we've seen. David Brooks supplies the aphorism for all of it — when intelligence is plentiful, volition is valuable — and Stanford's updated labor data shows the market is already pricing that scarcity.
News
The University of Chicago's social-sciences Core sequences will be taught as “analog” courses this fall: paper texts, device-free discussion, and no AI tools for students or instructors. A memo from divisional master Jenny Trinitapoli, obtained by the Chicago Maroon, reports a “strong consensus… that our sequences are best understood as a pedagogical setting without AI” and argues the Core should “first equip students to read, write, and think without dependence on AI.” All syllabi will be written by human scholars; AI-assisted grading “has no place in the SOSC Core” except to test it against human grading. Exceptions cover sanctioned experimentation, datasets and code, and disability accommodations. The timing is the story: the same September, every student gains free Claude Enterprise access under the university's new Anthropic partnership — access, a spokesperson noted, doesn't mean every use is permitted.
The interesting move isn't the ban — it's the unit of decision. Chicago set policy at the level of a whole first-year sequence, by consensus of the people who teach it, rather than leaving each instructor to improvise a syllabus paragraph. Our common-core and signature courses could do the same: decide together, per sequence, where AI belongs, so students meet one coherent norm instead of seven contradictory ones.
Source: "Sosc Core to Institute AI Ban, Technology-Free Classrooms This Fall" — The Chicago Maroon
Policy
MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training reported on August 13 after five months of listening sessions, and it reads as the constructive counterpart to Chicago's retreat: “This report is a call to action.” The diagnosis is blunt — AI is disrupting problem sets, take-home exams, and office hours, increasing isolation, and “eroding the ‘social contract’” between instructors and students, while detection software is unreliable and policing corrodes trust. In MIT's fall survey, over two-thirds of students said AI would matter to their careers, but “only 25% felt that MIT was adequately preparing them to use AI.” The committee's “pro-learner” recommendations: backward-design courses, shift toward oral exams and portfolios, require structured in-person social learning in every subject, put a reasoned AI policy in every syllabus, and never let AI be a thesis co-author.
Most of MIT's machinery — AI Leads in every department, implementation teams, a standing committee — exists to solve a coordination problem a school our size doesn't have. The immediately adoptable piece is the syllabus rule: every course states its AI policy with the pedagogical rationale, not just the prohibition. That single requirement forces the useful conversation — what is the cognitive work of this course, and does AI support or bypass it?
Sources: Report of MIT's Ad Hoc Committee on AI Use in Teaching, Learning, and Research Training · spotted via @Afinetheorem on X
Teaching
Dana Goldstein's New York Times feature assembles the case that writing's difficulty is precisely its value — it builds working memory, planning, and metacognition, or as cognitive psychologist Ronald T. Kellogg puts it, “Writing is a technology for thinking.” With surveys putting student AI use between two-thirds and 90 percent, educators are returning to blue books and proctored in-class essays; tech leaders counter that withholding AI is the real malpractice, with Grammarly's education chief calling this a “burn it down” moment. An MIT study found AI-assisted writers showed lower brain activity and struggled to remember what they had written. The middle path, from a UC Irvine expert panel: students draft their own work, with AI reserved for generating prompts and pre-submission feedback. A 17-year-old supplies the epigraph for the season: hard texts “are meant to be hard.”
The University of Sydney's “two-lane” model quoted here — some assessments secured and in-person, others explicitly requiring AI proficiency — is the cleanest assessment architecture we've seen for a writing-intensive curriculum. Worth mapping onto our signature-work sequence now: which lane is each assignment in, and do students know? Ambiguity, not AI, is what actually erodes integrity.
Sources: "What Do Students Lose When They Stop Writing?" — The New York Times (gift link) · syndicated copy at The Philadelphia Inquirer
Practice
Rod J. Naquin — an English teacher back in a Louisiana high-school classroom after seven years in educational leadership — has been publishing the most candid practitioner's audit we've read. His confessed mistake: introducing AI feedback before human feedback. “The technical execution worked great. The pedagogical judgment was wrong” — students experienced accurate machine comments as hollow until a person had responded first. What held up: grounding AI in real course materials (full novels, curriculum documents, rubrics) rather than asking what the model “knows,” and his insistence that there is no standalone AI literacy — “Using AI well for writing requires knowing how to write.” On X this week he pushed back on the much-circulated rule that students should use AI only for what they already do well, noting it is effectively what district policies already say — the hard part is designing assignments around it.
Two rules here transfer directly to our seminars. First, sequence matters: in small classes our comparative advantage is that a human reads student work first — AI feedback, if used, extends rather than replaces that. Second, Naquin's diagnostic question — does this AI use support or bypass the intended cognitive work? — is a better unit of course policy than any tool list, and small enough to fit on one syllabus line.
Sources: "What I got wrong (and right) about AI in schools" — Rod J. Naquin, The Science of Dialogue · spotted via @rodjnaquin on X
Ideas
David Brooks's Atlantic essay “The People Who Will Thrive in the AI Age” has become the summer's most-quoted framing of the human question underneath all the policy above. His guiding principle: “When intelligence is plentiful, volition is valuable.” What will differentiate people, he argues, is not how smart they are but “their relationship to mental effort” — drawing on the psychology of “need for cognition,” the spectrum from people who relish difficult games and dense books to the “cognitive misers” who avoid hard thinking when they can. The trait “correlates with intelligence but is not the same thing.” The thriving group, on his account, won't be those who passively use AI to work less, but those who wrestle with it to build their own capabilities. Critics note it's a values argument, not a data-driven one — which may be exactly why it travels.
This is vocabulary worth borrowing for orientation week and advising: the goal of a liberal-arts education in an AI economy is cultivating the appetite for cognitive effort, not stockpiling information a model already has. If Brooks is right that need for cognition is trainable and context-sensitive, then course design — how much productive difficulty we deliberately preserve — is character formation, not just content delivery.
Sources: "The People Who Will Thrive in the AI Age" — David Brooks, The Atlantic · unpaywalled copy
Labor market
Erik Brynjolfsson, Bharat Chandar, and Ruyu Chen have released a revised version of “Canaries in the Coal Mine?”, the Stanford Digital Economy Lab study tracking AI's employment effects through ADP payroll data on millions of U.S. workers, now updated through June 2026. The headline holds: “no evidence of widespread, economy-wide job displacement.” But the exception has grown — employment of workers aged 22–25 in AI-exposed occupations is now “19% below where it would be had it kept pace” with less-exposed peers, a gap that “has widened steadily” since first documented in August 2025. The mechanism is reduced hiring rather than layoffs, concentrated where AI substitutes for human tasks; where AI complements workers, employment is flat or rising. The authors stress these are descriptive early indicators, not causal estimates.
Our current students graduate directly into this market: entry-level rungs are thinning precisely where AI substitutes for the tasks new graduates used to be hired to do. The actionable half of the finding is the complementarity side — advising and career services should be steering students toward roles where AI amplifies judgment rather than replaces routine output, which is the same skill set items 3, 4, and 5 are protecting in the classroom.
Sources: "Canaries in the Coal Mine? Six Facts about the Recent Employment Effects of Artificial Intelligence" — Stanford Digital Economy Lab · spotted via @erikbryn on X
We are collecting examples for AI experimentation at DKU.
— Yisu