Ven Popov

Last year I was studying Group Theory just for fun but I never saw any use for it in my math psych work. Until now, which to my surprise proved not only useful but crucial in a problem I’m working on. when kids ask “when am I going to use any of this” in class the proper response is “you never know”

A preview of something fun coming up soonish - when are signal detection models or sets of random variables shift-representable?

YouTube video by 3Blue1Brown The most beautiful formula not enough people understand

Despite the already high bar, this brand new video by @3blue1brown.com is my favorite to date by far. What a beautiful lesson in building curiosity and intuition.

Not sure if that's the intended answer on this Captcha "Pick all creatures that can shelter in this item", but it's the one I'm going with

"What crisis? It's always been this bad! Stop the unhelpful criticism!"

What a strange take. Hopefully it will become less strange as I read the actual full paper.

Might be time for my 4th rewatch of BG. Every 5 years is a good rhythm

First time? Lucky you. I wish I could watch it for the first time. Though with my poor memory and the fact that it’s been at least 5 years since my 3rd rewatch, it might be time to do another

My favorite similar experience is when I hold onto an idea long enough to write it down only to find myself thinking the next day "I swear this made sense yesterday".

Very neat. I've been looking for an easy way to do the same in Quarto posts and this seems like the perfect solution.

My favorite move is the initial preparatory message "I will try to get back to you in two weeks after..." only to follow up 3 hours later with "nvm, i'm done". Much better than the inverse which I used to do. So now the wait is either expected or is replaced with a nice surprise :)

Anecdotally they are widely spread. There's also been some recent surveys published in Nature that show a widespread mental-health crisis especially among postdocs and junior faculty, who are also disproportionally burdeoned by review request just by numbers alone (many fewer senior faculty)

Why do people ghost referee requests? I’m sure there are other reasons, but burn-out, stress and depression are common ones (speaking from experience).

The Edge of Mathematics Terence Tao, the legendary mathematician, explains the promise of generative AI.

Also www.theatlantic.com/technology/2...

As another professor with ADHD and way too many commitments, I have implicitly trained all my collaborators that my response time is a bi-modal distribution of "I get excited and drop everything to do it right away" or, you know, the other extreme.

This experience also stresses how important it is even more than before to learn proper tools for collaborating on code and version control. In my case the entire project has a github repository, so the entire history is traceable and documented both for transparency, but also for safety

From the math community on Reddit Explore this post and more from the math community

And this is just one small part. It helped me formalize some of the ideas which I bounced back and forth as a conversation. And other colleagues are seeing similar results. Here's just one example from a post yesterday by a mathematician on reddit: www.reddit.com/r/math/comme...

For example, section 7 of this ongoing draft was entirely generated by Claude Opus 4.6 agent in a single prompt in VS Code copillot: venpopov.com/logit-probit...

It designed the questions, the simulations, ran the code, reviewed the output and iterated until ready. And I *agree* with its decisions

But more importantly, this intellectual work it does also enriches me as a scientist - it exposes me to ideas and techniques I was not aware of, even though they might be standard fare in other fields. We are way beyond the stochastic parrots of 2023.

Don't get me wrong. It's not perfect. It still needs steering and well-defined atomic scope. It becomes confused when the scope exceeds a certain threshold. But that threshold has already moved dramatically even relative to 6 months ago - I can reliably get it to do solid intellectual work

And yes, I check every result thoroughly, because it is my responsibility to ensure correctness. What we wrote in 2023 - that soon we'll be able to treat LLMs as research collaborators rather than computational tools, has for me undeniably come to pass

www.pnas.org/doi/10.1073/...

Claude Opus 4.6 is a surprisingly huge jump for a .1 upgrade. I've been astonished by how independently it can work and in the past month it has substantially helped me in projects in a way that I would absolutely think deserves co-authorship if it were a human collaborator.

School kids in the past: "The dog ate my homework"
School kids in the future: "The robot deleted my homework"

Ah, the cyberpunk version of "the dog ate my homework".

a man with a mustache is typing on a keyboard in front of a computer monitor ALT: a man with a mustache is typing on a keyboard in front of a computer monitor

should be safe now

that joke was so bad I heard future students groaning

and if I take the median rather than the mean, you could say I'm pretty *mad* about tibbles!

when I like them, I really like them, and when I don't, I really don't, so I guess on average I also don't have a strong opinion 😄

I'll shoot him a quick email

yeah, he seems like a serious person so likely a mishap

Also a bit concerning - can't find one of the only two refs:
"Some results on first-passage times of simple random walks"

Google scholar has no results for the title, and no such paper in that journal issue (nor authors) www.sciencedirect.com/journal/stoc...

Smells like LLM hallucination