The upside of giving talks (yeah, yeah, dissemination, collaboration etc.): 7 years later, I finally have my own CASA mug!

Ja.

Yes, it is a bit insane when you try to explain it to people from other fields. We finally did manage to get an "update": they can't get a hold of the editor, so why don't we please help them and send an email as well...

Whenever I talk broadly about research topics, I will catch myself saying "I dislike the distinction between pure and applied math...." and then have to explain that actually no, I am fine with it but reject the often implicit statement that one is better than the other.

The obvious addendum to this is that it is most likely also the way grad students—probably uniformly applicable—feel hearing me talk about joint work. And this direction is far more likely to be accurate/valid.

Indeed, reading the first few lines, the only thing I could think of was “this is like listening to a talk by your grad student” (albeit not uniformly applicable)

Non-issue as in any math/cs/... dept will likely see more egregious "style crimes" any given day of the week. Similarly to you, I have also been asked about wearing a suit, when at the same time mid-winter others were walking around in shorts and sandals.

I am entirely in the suits/jackets camp. I couldn't do the boots myself but given what people—I almost wrote "the kids"—are wearing these days, having cowboy-ish boots with suits, or generally with more-than-average formal clothes, seems like it should be a non-issue.

Took less than 10min from the airport and we’re in the gridlock of all gridlocks. Good to see the traffic people were warning me about 10 years ago has *not* improved

As an example where this is useful, we consider Schrödinger bridges where the reference measure comes from that of a (scaled) reflected Brownian motion. It remains an open question what types of sequences of cost functions, or equiv. reference measures, one can consider and still have an LDP.

entropic optimal transport. Specifically, we extend their results to cover also cases where you have a sequence of cost functions that converge in the right way (here, uniformly) to a limiting cost function. It then turns out that an LDP holds for the corresponding sequence of Schrödinger bridges.

Even managed to get a picture of four generations of the academic tree. Lots of large deviations and control-ish work in one photo.