<unnamed>

Public Key

npub1umssnfgeht7fvj26zxkfa6q7xadfqu2y5t4tpww37cqxt4dehafsq57ra6

Profile Code

nprofile1qqswdcgf55vm4lykf9dprty7aq0rwk5sw9z2964sh8glvqr96kum75cpz3mhxue69uhhyetvv9ujuerpd46hxtnfduqs6amnwvaz7tmwdaejumr0dsf07fgn

Publishing to

relay.damus.io

nos.lol

Last Notes

2026-08-19 23:26:20 CEST

- reply

Thanks! I stand corrected.

2026-08-19 18:20:13 CEST

- reply

I think I know the incident you are talking about and it wasn't an impossible test, the test questions even include good hints, but the LLMs often do something else weird instead. Specially under the incomplete sand-boxing and very lax monitoring that seems to have been put in place. (This whole comment is only a minor quibble about the word "impossible", I don't disagree that something should be done to avoid these problems in the future.)
This is the video I saw about it:
https://youtu.be/UyIiAIif5R0?si=6vsNfswEa7yVsIMt

2026-08-19 17:56:35 CEST

- reply

Oh, I didn't think you meant font rendering, partly because you said "font" without "rendering", but mainly because I don't understand why something Microsoft did (the "increased clue") would improve font rendering in LibreOffice.

2026-08-19 16:46:44 CEST

- reply

Users can choose fonts. Maybe neither software has improved and you just got a file from someone with good taste?

2026-08-15 18:40:13 CEST

- reply

Why in ten years? I think it already happened when I was in the PhD, like 10 years ago.

2026-08-07 02:21:41 CEST

- reply

A person called Jonathan Kraut wrote an assembler called Sassy in a standard called R5RS of a programming language called Scheme.

2026-08-06 17:33:27 CEST

- reply

Did you write "A person called Jonathan Kraut" instead of just "Jonathan Kraut" because you were worried people would think Jonathan Kraut was an LLM agent?

2026-07-28 00:43:30 CEST

- reply

I thought it was self-deprecating. As you pointed out, "gauche" means "left" in French, but it is also a word in English, that means "socially awkward, tactless, or rude". [@ramin_hal9001](https://fe.disroot.org/users/ramin_hal9001 )

2026-07-23 21:20:02 CEST

- reply

Shing-Tung Yau won a Fields medal in 1982 and is also Chinese.

2026-07-20 14:58:26 CEST

- reply

Sorry, no, I just saw those two links specifically on Hacker News.

2026-07-20 07:29:17 CEST

- reply

Claude Fable generated text indicating surprise too:
https://cdn.xcancel.com/pic/orig/EA1E99C3DAE99/media%2FHNpQXPsWoAAvXPH.png
https://cdn.xcancel.com/pic/orig/3B5C0B867A7C3/media%2FHNpQUaiWsAA43GQ.png

2026-07-20 06:51:44 CEST

- reply

Surprisingly low degree, right? Wow.

2026-07-07 22:44:19 CEST

- reply

I'm one of the organizers for the colloquium in my department and I've noticed that when we invite algebraists to give a talk, they very rarely give either examples or motivation! Well, maybe it's more accurate to say that they do give some brief motivation, but it doesn't convince me. Their motivation is like: "so-and-so study this variety of algebras that satisfy these identities. What happens if we change that sign or that exponent from 2 to 3? It's a whole new world!"
The talks themselves tend to be pretty clear and easy to follow. I always understand what they did, but almost never have any idea why they did it, why they think it was interesting. I've started giving algebraists extra instructions before their colloquium talks. It's definitely the broad field of mathematics in which I'm most careful about who we invite.

2026-07-07 04:45:37 CEST

- reply

Minor nitpick it's only "soccer" in Canada and the US, not so in the second largest country in North America (by population), where we say "fútbol" like you do in your blog. 🙂
EDIT: I just learned from Wikipedia that while North America does include all of Central America, there is also a term I had never heard before, "Northern America", that you can use for the parts of the Americas that are north of Mexico, the biggest chunks of which are the US and Canada, of course.
https://en.wikipedia.org/wiki/Northern_America

2026-07-07 04:30:55 CEST

- reply

Sorry to go off on a tangent, but is football really called "footy" in Canada? I've only ever heard the word "footy" used to refer to Australian Rules Football¹, which is played with with an oval ball (similar to the balls used for rugby or American football). I apologize but I thought football was called soccer in Canada, like it is in the US.
¹ https://en.wikipedia.org/wiki/Australian_rules_football

2026-06-13 18:44:10 CEST

A student who puts their name in the file name of a PDF they send me instead of just calling it thesis.pdf is clearly a genius and I almost don't need to read the thesis. 😛
#AcademicChatter

2026-06-11 21:18:59 CEST

Wow, this is bleak.
https://blog.avas.space/llm-circus/

2026-05-29 08:02:41 CEST

I learned a cool Euclidean geometry fact this week in our colloquium, in a talk by Benoît Bertrand of the University of Toulouse on joint work of his with my UNAM colleague Lucía López de Medrano. They don't think this geometry fact can be new, but also haven't been able to find a reference. The theorem is a generalization of the fact that the sum of the angles of a triangle is 180° to higher dimensions.
This generalization is not that the sum of the solid angles of a tetrahedron is constant: that's not true! If you think of tetrahedra that are almost flat, that is, whose vertices are nearly coplanar, there are two types: (1) the vertices are nearly the vertices of a convex planar quadrilateral, or (2) the vertices are nearly a triangle with a point inside it. In case (1) the solid angles are close to 0; in case (2) the one at the central vertex is close to half a sphere (2π steradians), and the others are close to 0. It turns out that those are the extremes and the sum of the solid angles of a tetrahedron is always between 0 and 2π steradians.
So how does their generalization go? Consider an n-dimensional simplex and select one of its vertices, say P. Now consider all ways coloring the vertices of the simplex with either red or blue in such a way that P is blue (this is half of all colorings). For each coloring look at the vectors of the form R-B where R is some red vertex and B is a blue one, and take the cone they generate (that is, the linear combinations of those vectors with non-negative coefficients). Their theorem says that those cones tile a half-space! (Here "tile" means the interiors are disjoint and the union of the cones is a half-space.)
Next, a proof. 1/2

2026-05-29 05:17:22 CEST

- reply

Ooh, a conspiracy theory I can get behind! Now I want to measure the Erdős number of these people with and without that paper, 31 data points is enough to make a nice plot!

2026-05-28 18:40:16 CEST

- reply

I don't think it has a built-in dictionary, aren't those just built-in commands to access external (non-built-in) dictionaries?

2026-05-24 23:02:01 CEST

Woah, a geometric topology paper with 31 authors!
It's titled "Algorithms in 4-Manifold Topology", and they show that there exists an algorithm that takes as input two closed, simply connected, topological 4-manifolds and decides whether or not the manifolds are homeomorphic.
https://arxiv.org/abs/2411.08775
(Ah, the authors are all the participants of a workshop on algorithms in 4-manifold topology! That makes sense.)

2026-05-14 16:39:22 CEST

- reply

I've now listened to the episode, it was quite fun! I definitely should play around with some Lisp Machine emulator to experience these "presentations" for data first hand. I think Emacs can probably emulate some (but most likely not all) of that stuff through judicious use of text properties.
One minor point outside the main topic: [@ramin_hal9001](https://fe.disroot.org/users/ramin_hal9001 ) mentioned that he found Hyperbole orthogonal to Embark since Embark focuses on offering you a bunch of actions, while Hyperbole focuses on "doing what you mean" in context. I wonder [@ramin_hal9001](https://fe.disroot.org/users/ramin_hal9001 ), have you tried the embark-dwim command? I don't see any obstacle to get it to do whatever Hyperbole's action key does at any given point. Out of the box it already does the same thing as the Hyperbole action key in some places —for example, it'll evaluate an s-expression if you are on the opening or closing parenthesis and it'll follow URLs or, more generally, any org link (even in non-org buffers).
[@kentpitman](https://climatejustice.social/@kentpitman ) @npub1fp0…54p5

2026-05-14 15:26:43 CEST

- reply

I think it was Tony Hoare, and the quote is: "Here is a language so far ahead of its time that it was not only an improvement on its predecessors but also on nearly all its successors."
Also, I believe it was specifically about ALGOL 60.
@npub1fp0…54p5 @npub1vlc…mzqq

2026-05-14 02:38:44 CEST

- reply

hahaha, after attempting to listen to this episode I remembered why I am not already subscribed to this podcast, which clearly interests me! It's because I've always had trouble playing the episodes! My current podcast app can subscribe to the feed and can download the episodes but cannot play them at all and just gives me an error message! I think in my previous attempt some months ago I could not subscribe to the RSS feed for some reason, but I had managed to download some sort of audio files that I could play but could not rewind nor change the playback speed of (and maybe it wouldn't even let me pause?, I can't remember). I have never ever had any trouble subscribing to or listening to any other podcast, it's just this one that is insanely difficult for me for some reason. 😅
@npub1hll…0fjp @npub1fp0…54p5 [@kentpitman](https://climatejustice.social/@kentpitman ) [@ramin_hal9001](https://fe.disroot.org/users/ramin_hal9001 )

2026-05-13 04:41:21 CEST

- reply

That sounds interesting. I'll listen to the podcast to see what you were talking about. At first blush, this talk of surfacing reminds me of embark-eldoc-first-target, which will display the target and point and its type in the echo area.
@npub1fp0…54p5 @npub1yna…cu6l [@kentpitman](https://climatejustice.social/@kentpitman ) [@ramin_hal9001](https://fe.disroot.org/users/ramin_hal9001 )

2026-04-16 23:58:35 CEST

- reply

I found this paper that seems to take the history seriously and I would believe the author did read Liouville directly: https://doi.org/10.2307/2689612.
It seems that Liouville didn't use the term "elementary function", but rather "explicit finite function". His explicit finite functions do however correspond to the broader notion that is closed under algebraic extensions, not just the "folk elementary functions". Also, the author, believes (like I did) that Liouville's class of functions is what we now call "elementary functions". Funny.
I guess since Liouville's purpose was to prove that certain functions don't have elementary integrals (well, explicit finite integrals), he went with the broader notion to get a stronger theorem!

2026-04-16 23:53:37 CEST

- reply

You're right, I hadn't seen that, thanks!

2026-04-16 22:54:28 CEST

- reply

That's interesting. Maybe it's just not true that Liouville gave the other definition. It could be that people misattribute a much more modern definition to Liouville.

2026-04-16 04:54:42 CEST

- reply

Really? I thought people very carefully said that the quintic is "not solvable with radicals", and that the original definition of elementary functions given by Liouvile already had them closed under solving polynomial equations, logarithms and exponential (that is, I thought the differential algebra definition @npub1nf4…nqe4 explained elsewhere in this thread is just a modern restatement of Liouville's original definition).
Is this other definition of elementary function you guys are using supposed to be older or more recent than Liouville's?

2026-04-04 01:08:10 CEST

- reply

Turkey is much lower on the list than I expected!

2026-03-27 15:41:18 CET

Man, LLMs really drive home the lesson that people differ *greatly* in how much they care whether the statements they make are accurate.

2026-03-10 14:39:10 CET

RE: https://mstdn.social/@rysiek/116204163611415275
It looks like the International Congress of Mathematicians is less serious than the Ig Nobel Prize. 🙁
#note1cxg…d0r7

2026-02-25 20:42:31 CET

Nice plot summary... Google just keeps getting dumber.
https://media.mathstodon.xyz/media_attachments/files/116/132/973/702/318/690/original/06c5842fe01fb991.png

2026-02-21 22:49:42 CET

You've probably heard of the Birthday Paradox, that among 23 people the probability that two share a birthday is more than 50%.
This is easy to prove by calculating the complementary that no two of the 23 share a birthday. There are 365 options for the first birthday, then only 364 for the second, 363 for the third, etc. So the probability that all 23 people have different birthdays is (365×364×...×343)/365²³ ≈ 0.4927.
That calculation assumes the uniform probability distribution on birthdays, that is, that every day is equally likely to be a birthday. That's not a very realistic assumption for the actual distribution of people's birthdays, so it's worth thinking about what happens for other distributions. My instinct was that any non-uniform distribution has some higher probability clump, where birthdays would tend to bunch up with even fewer people; so that any non-uniform distribution is even "more paradoxical" than the uniform one.
This turned out to be easy to prove.
1/2
#math #probability

2026-02-02 01:26:06 CET

- reply

After reading what your wife thought, I took a look at some reviews and it gave me flashbacks of reading my student evaluations: one review sides with your wife on the need for editing, calling it "a tedious, bloated epic"; another review calls it "tightly constructed". When I taught undergrads in the US, I remember getting student evaluations, from the same class, one which said I was "the best math teacher I've ever had" and one which said "I don't know how the university lets him teach".

2026-01-30 22:38:34 CET

- reply

Hey! @npub19ua…53cv is a coauthor!

2026-01-29 06:10:19 CET

- reply

Have you also tried Phanpy on your phone (or only on the desktop)? I love it there too. If you did try it, what made you go with Tusky instead?

2026-01-25 18:19:35 CET

There is much truth to this post about about #math textbooks.
https://media.mathstodon.xyz/media_attachments/files/115/956/870/326/807/020/original/8abb41b70b5f11cc.png

2026-01-22 05:07:21 CET

- reply

How do you format the links like that? Are you posting from a non-mastodon client? Does that mean Mastodon clients can display links and other html formatting but have no way of producing it? That's funny.

2026-01-16 20:19:42 CET

- reply

Maybe that's the key! Maybe the methodology is that you compute some global "humanities score" for each university and then "normalize" by dividing by the number of humanities departments. 😛

2026-01-16 17:50:40 CET

- reply

I agree that the MIT claim is very suspicious, a little funny too.

2026-01-16 17:34:41 CET

- reply

On my phone it lets me scroll the list to see past the fourth one. I haven't tried on a desktop computer yet.

2026-01-14 15:38:37 CET

- reply

What was your use case for regexps?

2025-12-23 19:18:04 CET

- reply

It's probably just a typo for "among us", the popular videogame that features characters that look like that shape. @npub1zz2…9mhx

2025-11-18 16:28:43 CET

- reply

I wanted to say this, thanks for saving me the effort. (Except I would have said "we" since I only sound like an American.)

2025-09-13 12:58:13 CEST

- reply

Will do.

2025-09-13 01:45:29 CEST

I can't read the nLab anymore! A few days ago it started "verifying" I was human. Each verification took around 30 seconds, which was annoying (a very human emotion!), but at least I passed them! Today I can't get past the verification screen, which makes me sad (another human emotion).
EDIT: I rejoice for the nLab deems me human once again! (Rejoicing also seems pretty human to me.)

2025-09-10 13:27:56 CEST

- reply

Beautiful!

2025-09-04 18:25:21 CEST

- reply

I thought of it slightly differently, instead of mapping the triangle to an equilateral one, I thought of mapping it to itself but permuting the vertices.

2025-08-30 05:16:15 CEST

- reply

Someone printed one! https://mathstodon.xyz/@fractalkitty/115115418786961240

2025-08-11 09:17:59 CEST

- reply

This matches what I was planning pretty closely (and the usual focus on just irreducible varieties also bothers me, as does the similar focus on connected covering spaces, for example). I mentioned you and Dolan because I was thinking of that dimensional analysis/doctrine stuff and wondering if there was a more systematic write up of it now.

2025-08-09 23:15:49 CEST

- reply

I'm particularly interested in what you and James Dolan would suggest, @npub1nf4…nqe4

2025-08-09 23:14:46 CEST

The algebraic geometers at my university were all distracted at the same time and I got assigned the core algebraic geometry graduate course while they weren't looking! What should I endeavor to do to this generation of Master's students?

2025-07-31 16:36:51 CEST

- reply

Arguably this stick insect is both! Also, they've changed illusive to elusive on the website already, I only see illusive in the link preview here on Mastodon.

2025-07-30 19:29:33 CEST

- reply

MEXICO - Pedro Páramo (Juan Rulfo, 1955)

2025-07-21 04:33:44 CEST

You know how people advise you not to use mathematical formulas in section titles in LaTeX? Listen to them!
https://media.mathstodon.xyz/media_attachments/files/114/888/870/407/228/144/original/61e794b3e6e090c9.jpg

2025-07-18 06:16:15 CEST

- reply

Like the Google Buzz trick from Doghouse Diaries (http://thedoghousediaries.com/1267):
https://media.mathstodon.xyz/media_attachments/files/114/872/273/256/613/662/original/066f90a2b68dca17.png

2025-07-08 00:38:49 CEST

- reply

It bothers me a little bit that it has both "National" and "of Athens" in the name.

2025-07-07 21:00:52 CEST

- reply

This is about grouped notifications, that is, seeing notifications for a single toot group together in a single list: instead of a separate notification for each time someone favorites or boosted a given toot, you see a single notification per toot listing all the people that favorited or boosted it.
This could of course be implemented in each Mastodon client separately, but Mastodon decided to do it serverside for efficiency and uniformity. On of the Mastodon clients I use, phanpy, has always shown grouped notifications (at least since I use it); but another client I use, @mousebot's mastodon.el, does not show notifications grouped by default. And today I learned it does have a setting to group them!
Is that what you wanted to know?

2025-07-07 18:34:21 CEST

Oh my god, TIL about mastodon-group-notifications! Now #mastodonEl is perfect! (Or rather: it already was perfect, I just didn't know it was.)

2025-07-02 07:30:24 CEST

- reply

Hahaha, I'm sorry, specially if I just cost you addition to heaven. 😅

2025-07-02 07:21:32 CEST

- reply

This gives new meaning and new higher stakes to the phrase "cinnamon challenge".

2025-06-05 19:21:19 CEST

- reply

Oh, that's right, getting the appropriate link is a little awkward with just Org. I should have mentioned what I use for that: org-ql-find together with embark. Org-ql has several flexible command for searching for heading in an org file, org-ql-find in particular will prompt you in the minibuffer for some search terms in a simply syntax, and then it lists the matching heading as completion candidates. (It's mean to be used with a completion UI that shows the completion and lets you navigate among them, a completion UI such as icomplete, vertico or ivy.) The embark package let's you act on the completion candidates and in particular comes with an action called embark-org-insert-link-to, which inserts a link to the currently selected heading. I bind org-ql-find to M-s q and embark-act to C-., and embark binds embark-org-insert-link-to to j, so altogether I insert a link to a heading in the same org file as follows:
1) M-s q search terms
2) arrow keys to select the correct heading in the minibuffer
3) C-. j
It's very fast and convenient. You could also replace org-ql-find with consult-org-heading, if you prefer, but I like that org-ql-find does full text search in the entire body of the heading by default.

2025-06-05 17:34:50 CEST

- reply

Yes, just use Org on its own!

2025-05-23 15:48:53 CEST

- reply

Damn autocorrect! Thanks for pointing out the typo.
Scotland? I somehow thought you had left the US for Singapore. At any rate, I'm glad you left the US.

2025-05-23 14:34:33 CEST

- reply

I think the article make a pretty convincing case that the brain drastic is already underway, it's funny they called it a "danger" in the title.

2025-05-08 02:20:42 CEST

- reply

The six digit numbers whose first three digits are the same as the last three are the multiples of 1001 (because abc * 1001 = abcabc). Since 999999 is a multiple of 1001, being a multiple of 1001 is well-defined modulo 999999, i.e., if a = b (mod 999999) then a = b (mod 1001). Finally, since 123123 is a multiple of 1001, it doesn't matter what the tower of fives is, the product will be a multiple of 1001.

2025-05-05 14:31:32 CEST

- reply

The author of that article, @npub1hr7…jxyc, was written plenty of sensible stuff on AI including a book, The Intelligence Illusion.

2025-03-17 04:17:56 CET

- reply

Because of stories like these I cancelled my trips to the US for math conferences for the immediate future. I'm sorry I won't get to see many American friends of mine, but the minute I sent those emails cancelling, I felt relieved.

2025-03-15 16:31:01 CET

- reply

The partial automorphism you found are inner on the a=1 subgroup, but they might be non inner once you extend them to the whole group.

2025-01-10 13:15:11 CET

- reply

When I visited Copenhagen I got a public transit card at a machine at the airport. The machine asked me what felt like a very long series of questions in Danish (there was probably an option somewhere to switch to English which I missed). I clicked on buttons somewhat randomly and got a card, so I considered the operation a success. Later I found out that I had obtained a transit card for dogs —which apparently are (1) allowed on public transit, but (2) have to pay!
Now, dog was only the default for the card, mind you, so I could still use it: everytime I got on a bus or train I just had to add one human and remove one dog to my travel order, a little game I called "Copenhagen: Become Human".

2024-11-20 12:37:40 CET

- reply

I see, thanks. That makes sense.

2024-11-20 05:40:09 CET

- reply

This thread was pretty interesting and I enjoyed the article of Doctorow's you linked to, but I think I may have missed something: you explained why you aren't moving to bluesky, but did you also say why you are bothering to mirror your posts there?

2024-10-06 02:44:34 CEST

- reply

I guess the answer is yes and a little bit no. There is the Risch algorithm, which does decide whether or not there is an elementary anti-derivative, but it uses as a subroutine a function that tells you whether two constants are equal. Such a subroutine is not know to exist for constants you can write in terms of rational numbers and elementary functions, and I think that if you throw in absolute value it is in fact known to not exist! So, modulo it being very difficult to tell whether a big expression that evaluates to a number is zero or not, you can decide whether or not elementary functions have elementary anti-derivatives. Personally I think "compute the constant to 500 decimal places" is probably correct enough of the time to not consider this a problem. The Risch algorithm is also extremely complicated and I seem to remember either no computar algebra system implements it completely or maybe just one or two do.

2024-10-04 08:49:44 CEST

- reply

Oh, you're right: this is exactly equivalent to asking which does are parallelizable and the classic answer is that only S¹, S³ and S⁷ are! (Though I only know how to prove that using K-theory which was forbidden by OP.)

2024-10-04 08:42:19 CEST

- reply

de Rham cohomology works half the time for this question! For SO(n) with n=2k+1 you have exterior generators in degrees 3, 7, ..., 4k-1; and the slightly different pattern you mentioned for even n is that if n=2k+2 then you have exterior generators in those same degrees 3, 7, ..., 4k-1 and an additional exterior generator in degree 2k+1. This means that when n is odd SO(n+1) has the same de Rham cohomology ring as SO(n)×S^n! (And when n is even the rings are non-isomorphic so the argument does work in those cases.)
When n=1 or 3, then SO(n+1) really is diffeomorphism to SO(n)×S^n, so I'm not sure what to expect for odd n.

2024-08-23 19:48:22 CEST

- reply

Y separando a los burgueses del proletariado.

2024-08-23 14:27:23 CEST

- reply

Or a misunderstanding? Every loop on the sphere is homotopic to one that does not hit every point on the sphere, maybe the lecturer meant something like that.

2024-08-23 13:59:00 CEST

- reply

But there are such loops! They are called space filling curves. The most popular examples are continuous functions [0,1]-->[0,1]^2 which hit ever point in the square but any such example is easily modified too obtain a loop on a sphere.
https://en.wikipedia.org/wiki/Space-filling_curve