Creative Force Dispatch: A Correct Solution the Student Can't Explain. What's it Worth?

New format: five reads and one question we answer together. Vote, tell us why, and next week's Dispatch opens with how you voted and the top comment.

Research mathematics just ran into a problem every math teacher knows. OpenAI's Navier–Stokes proof appears to be correct, but experts say it "doesn't teach us much." The first 3D "Einstein" tile was, in its finder's words, "found by Astra, not me." Every teacher has graded that paper: the right answer, with nobody home behind it. Harvard's PhD summit reaches for the teacher's fix: have students defend it out loud. The Khanmigo study shows the catch: students walk away when the tutor asks them to think.

THE FIVE

[Radio · 4 min] NPR: AI solved one of math's hardest problems. Humanity learned nothing (so far)

OpenAI's 166-page Navier–Stokes proof passes a computer check, yet the mathematicians reading it say they can't learn much: "The paper is not written for humans." It's the clearest case yet that a proof and an explanation are different things.

[News] Quanta: Software developer finds first 3D "einstein" tile

An Athens software developer used OpenAI's GPT-6 Astra to find Chair44, a single shape that fills space but never repeats. He says he understands the intuition, "not the mathematics of it." Amateur discovery changed the meaning, and so did checking: an expert had to cut the 61-page paper to eight pages so others could follow it.

[Report] Harvard CMSA: Recommendations of the Summit on PhD Math Education in the Age of AI

Twenty-four senior mathematicians propose that a PhD "should not be awarded primarily on the basis of the text of the dissertation," but on a defense that demands complete mastery. Read "dissertation" as "homework," and it's a blueprint for assessment in any math classroom. It's a first draft, and they want feedback.

[Research] Hechinger Report: Students didn't get answers from Khanmigo. They didn't want its questions, either

A two-year randomized trial in 18 Tennessee middle schools found that Khan Academy helped struggling students, but no more than it does without AI. When the Khanmigo tutor offered hints and questions instead of answers, students mostly stopped using it. Productive struggle has a motivation problem, and a Socratic chatbot doesn't solve it.

[From the archive · 1994] William Thurston: On Proof and Progress in Mathematics

Thirty-two years ago, a Fields medalist wrote this week's best reply: "We are not trying to meet some abstract production quota of definitions, theorems and proofs. The measure of our success is whether what we do enables people to understand and think more clearly and effectively about mathematics."

**Got a read that belongs in next week's five? Drop the link in the comments.**

THE QUESTION

A student hands in a correct solution they can't explain. What's it worth? Vote below, then tell us about the last time this happened in your classroom and what you did. Can't pick? Tell us what it depends on. (One Choice Each)

Full credit. Correct is correct.

Credit once they can explain it.

Partial credit. Finding it counts.

No credit. It isn't their work.

1 vote

3 comments

Creative Force Dispatch: A Correct Solution the Student Can't Explain. What's it Worth?