The new gen ed reforms (like at @UTAustin) will fail if they continue to think of AI solely as a subject to be taught rather than a technology that transforms the whole endeavor. Short version: the entire common intellectual experience of every @UTAustin graduate fits in one context window Everyone is applauding the return to Great Books. But the older and more central the text, the denser the AI training data around it. Mandating a small number of canonical texts and requiring faculty who may not have training in those texts to teach them is foolhardy and expensive in this era. How is it smart for a university to install as a foundation the corpus where AI is strongest? What a gift to AI firms! It will cost the student about $1.50 to use AI to research and write papers, so of course they will. This is obviously a back of the napkin calculation but it's the kind of analysis UT Austin might think of doing: Cohort: 9,900 first-time-in-college students entered UT Austin in fall 2025. Masterworks is 3 credits, required of all, so one cohort takes it once. Assumptions: 1.33 tokens per word of English prose; a semester syllabus meeting the three-mode rule at roughly 250,000 words — one novel at 120,000, one play at 25,000, poetry selections at 15,000, two further substantial texts at 90,000. The course corpus: 250,000 words × 1.33 = 332,500 tokens, call it 330,000. That is the entire semester's assigned reading as a single input — the whole common intellectual experience of every UT graduate, in one context window. One pass of the full corpus per student: 330,000 × 9,900 = 3.27 billion tokens. Writing: three essays of 1,500 words. Drafting in chat with pasted excerpts and four iterations runs about 20,000 input and 8,000 output per essay, so 84,000 tokens per student. 84,000 × 9,900 = 832 million tokens. Cohort total: about 4.1 billion tokens for one semester of one course. At $3 per million input tokens, 4.1 billion costs roughly $12,300. With prompt caching or a cheaper model, under $2,000. The complete works of all forty authors on the list come to about 21 million tokens. This is the way educators need to be thinking. Even just to say this calculation was run.