So apparently college students are one-shotting essays with LLMs and teachers are quitting because "they want more from life than grading an AI’s essays."

Erik Hoel's solution is to monitor the process of creating an essay rather than just judging the output. Not a terrible idea, but it strikes me as a pretty complex (and tech-dependent) thing to do.

We already have this technology though — it's called _talking_ to people. Socratic dialogue has been the gold standard of learning since forever, and is impossible to fake (particularly if you're in the room together).

At the small college I went to, we had "recitations" every Thursday/Friday with our professors where we talked about the week's readings. They were the best part of the week because we got to hear directly from our professor, and articulate for ourselves what we got out of the reading. The problem is that this approach doesn't scale — but that's part of what makes it worthwhile.