I would be grateful for school-level scaling of LLM management, if only to signal to students that they have to think about LLMs and respect different class policies. During my new faculty orientation, all the incoming APs could talk about was handling LLM use - but we're all on our own with it.

In general, I spent a lot of time at the start of the semester expectation-setting and discussing with the students why our particular course would not allow LLMs, and I've been pretty happy with how it's going.

This my approach for my undergraduate writing seminar - which is part of a Cornell series of classes explicitly supposed to teach students how to write. But it's definitely labor-intensive!

It's not like there won't be solutions. But it's a real interruption in the production of scientific knowledge.

Oh dear, do excuse my rounding. It's actually 83% of the course.

There are variations on the general theme.

"These pre-trends suggest that we have a time-varying confounder."

"That there is a discontinuity in the conditional expectation function of our potential outcomes under treatment."

"See that curve? No more linear functional form assumptions for us."

To be clear, I don't think they even realize the kitten has two homophonous names. We don't have a family group chat, so my sister and I text with each parent but our parents rarely text each other.

For a lot of people, intellectual labor *is* the troublesome labor.

I'm tempted to hustle over to a mall and start checking out tags.

Matt, it genuinely really shaped how I thought about estimand selection. I quoted that tweet on my quant methods qualifying exam in grad school. 😂