If only there were a country with a strong cultural connection to Canada that is also the world leading whisky producer, it would be possible to substitute a whole range of much better options.

They said they would stop the boats and instead reduced the number of international PGTs and people on work visas.

An under predicted POLAR5 was more likely to have been gated by asking rates. Opportunities to trade up in August are limited, so the incentive to reapply as PQ next cycle is relatively high. This is what we see in the data.

Green: pleasant surprise on results day! POLAR1 and POLAR5 students start in different positions, though. Contextual offers means that a POLAR1 would be able to apply more widely in the first place, and targets make AOs willing to do business, if you want to trade up.

Dark red: massive negative shock on results day. You probably missed your firm badly and insurance too. Clearing is there, but leaving the cycle without matriculating to reassess the whole plan is also an option.

Gray / light red: either no information or mildly negative information on results day. You very likely made either your firm or insurance offer, which now looks pretty attractive. Either way you take it. This is exactly how UCAS is supposed to match efficiently, and it shows in the figure.

Here's the main figure in the article. Let's go through the thought process in August for each color.

Watch this thread for more rigidity related conference news.

August: in a new one for me, I'm co-organising an invited track "Advances in Rigidity Theory, Multi-Agent Formations, and Distributed Localization" at the IFAC World Congress (a big control theory conference). I'm always excited to work with a new community!

May: ICMS has a workshop on "Rigid structures in algebraic combinatorics and algebraic statistics". Come check out this cool Scottish conference!

February: Timo de Wolff's group is organising a workshop on "Positivity, Convexity and Computation in Rigidity Theory". I'm really excited about this workshop, where I'll talk about the energy-based approach to higher order rigidity and prestress stability.

January: there's a Banff Chennai workshop on "Rigidity, Flexibility and Complexity of Geometric Constraint Systems". I've never been to India, and my postdoc Aliso la Porta will be presenting some new stuff, so generally exited about this one.

I think they work fine as a fancy notepad. Students who do their psets in notability and mark up the margins of the lecture notes still seem engaged enough not to worry me.

Tablets are fine. Laptops are not great. Phones are the absolute worst. If I could, I’d do tablets only.

Even well meaning hectoring is probably going to fall flat in the face of incentives and AI being pretty good. Just be uncool like we are in maths and do old-fashioned summative assessment. It works better, and you don't have to nag.

The way to keep your assessment clean is just to align it with what you want to teach. If it's mastery of the course packet and the ability to quickly structure a coherent argument, do in class essays. For polished writing they can defend, viva them to poke at things and discuss the refs.

Another issue, which is specific to the UK, is that degrees don't have a writing distribution requirement. A number of students haven't gotten a lot of direct instruction in writing, and they will use AI to punch up text and talk over structure. I think this is hard to stop.

The AIs of 2025 will produce a logically coherent, if slightly flat essay (get a coupon for ChatGPT+). It's not excellent work, but the target audience for this "don't use AI" document isn't producing excellent work and already knows that. The question they are asking is whether it's adequate.

My experience with misconduct on psets is that the kind of student who cheats: would be pretty happy to reliably get a II/2 on the module; and are willing to get help from others who are barely any better off. It's a different field, but I assume the dynamics are similar in essay subjects.