RE: https://mathstodon.xyz/@DavidWood/116983339289510925
Recieved by Annals of Mathematics in 2018 and accepted in 2026, just a few days ago!
1. I had no idea this very famous paper was not published.
2. I will never complain again about how long it takes me to get a paper rejected or accepted! (My record is 2 years —that one was rejected— but the average is much closer to 1 year than 2, let alone 8.)
«Amid the rise of artificial intelligence, technophobes and Luddites have continued to insist that machines “can’t really write”—at least not the way humans can. Those naysayers will be hard-pressed to wave away The Great Gatsby, the debut novel from the super-advanced Xerox 914 photocopier—an exciting new voice that wrote Gatsby after being trained on a data set comprising a paperback copy of F. Scott Fitzgerald’s The Great Gatsby.»
https://www.mcsweeneys.net/articles/book-review-the-great-gatsby-by-the-xerox-914-photocopier
I just read the transcript of @sacha and @karthink's conversation with John Wiegley about his #Emacs #OrgMode setup that he uses to organize everything in his life.
Wiegley is living in the future.
He has over forty thousand entries spread throughout more than three thousand org files, around 28000 TODO items with about a thousand of them open. It's too much data to even look at so he has some very sophisticated ways of dealing with it. His org agenda views will surface random selections of entries for him to review; upon review with a few keystrokes he can schedule them to work on or schedule them for further review in the future. He decided on a fairly strict set of conventions for how org entries should be written, including that the leading verb must be chosen from a controlled vocabulary and wrote a Haskell program to parse his org files and put them into a Postgres database which he can query. The database has vector search to find semantically similar items even if they are worded differently. He also has a local LLM backing an instance of OpenClaw that can query this database, access his org files and talk to him on Discord and Whatsapp, letting him find relevant entries from his org files by messaging the agent from his phone.
It's amazing, maybe not something I'd want for myself, but definitely a fascinating read!
I had heard of people clapping when the plane lands, but as far as I can remember I have never been on one of those flights before today. Man, is it weird. 😅
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. 😛
Wow, this is bleak.
A former postdoc of mine is applying to data science positions in industry now and has listed me as a reference. I haven't been contacted yet, but I am slightly worried I won't know how to give an effective reference for this kind of job: I only have experience with academic recommendations. I'd be grateful for any advice on how to do this!
This was an impulsive decision. I am now a cautionary tale.
I'm going to use it for a bit, despite complaining about things popping up automatically in the past.
At least I finally understand why it's called dog-fooding. 😅
My defense is "Prot made me do it", see https://www.youtube.com/live/xl-ifABU45A?si=ht8hTSWSAfvXQ5Bl&t=3366
Embark has a new embark-automatic-prefix-help-mode which behaves like which-key's default configuration: if you type a prefix of an #Emacs key binding, after a small (configurable) delay, you get reminded of what keys are under the prefix.
This functionality simply calls the existing embark-prefix-help-command on an idle timer. I've already written about that older functionality in my blog:
https://www.matem.unam.mx/~omar/apropos-emacs.html#the-case-against-which-key-a-polemic
but a quick summary is that if you use
(setq prefix-help-command #'embark-prefix-help-command)
then pressing C-h (or whatever you have set as help-char, if you've customized it) after typing a key binding prefix will prompt you with completion for a command under the prefix. If you use Vertico's grid display this looks a lot like which-key, but is more powerful because you can filter by typing bits of the name or of the rest of the keybinding, if you use Marginalia you can see first lines of docstrings, and you can use embark-act on the commands to do things like display the full help, go to the corresponding part of the info manual, jump to definition, etc.
So if you weren't using that just because you didn't want to press C-h and instead wanted it to happen automatically, the new embark-automatic-prefix-help-mode is for you. (I won't judge you if you use it!)
Oh, it turns out the person wrote the code: "the LLM did not write any code, it updated the README and copy-pasted my code then added whitespace, comments, and function documentation". That changes the situation, right? I can review this like any other human-written PR?
Ooh, got my first LLM-co"authored" pull request! Luckily, I think I can just turn it down saying GNU Emacs cannot take that contribution without getting into a personal discussion.
Oh, I almost forgot! Exercise: show that for a triangle this implies the classical theorem on the sum of angles of a triangle.
Here goes my proof of their result, which is not exactly the same as the proof Bertrand gave in his talk (that one inducts on the dimension):
First of all, the points of these cones are what are called "affine combinations" of the vertices of the simplex, that is, linear combinations whose coefficents sum to 0. And the vertices of a simplex are affinely independent, which means you can talk about the coefficient of a vertex in such a combination and it is uniquely determined.
Now those cones have disjoint interiors because, given an interior point X of one of the cones, you can figure out exactly which one it is: the cone corresponding to the coloring whose blue points are the vertices with a negative coefficient in X!
Finally, the union of the cones consists of all affine combinations of the vertices of the simplex in which the coefficient of the special vertex P is negative. That's a half-space! in the coordinates given by these coefficients it is given by the system x₁ + x₂ + ... + xₙ = 0, x₁ < 0.
2/2
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
Testing strategies to remove any sensitive information from my #OrgMode agenda and org-ql search results for tomorrow's #Emacs Chat with @sacha.
I think just removing the sensitive org files from the list of files I usually search (appropriately named org-ql-usual-files) leaves enough search results to comfortably demo things (and removing those might be a neat 20 second demo of Embark). I might add some large extra file, to get more search results, like https://doc.norang.ca/org-mode.org
Thanks to GitHub user aikrahguzar for telling me that #Emacs 30 added support for outline-minor-mode to eww (and more generally to shr output)! This lead me to learn about the outline-search-function variable that can be used to add support for outlines not based on regexps. I now updated Embark to also recognize that type of heading in general, and specifically in eww buffers. (The outline-search-function variable has apparently been there since Emacs 22, so I have no excuse for not finding it when I initially added support for outline heading to Embark. 😅)
[And on a personal level, aikrahguzar's information also allowed me to delete an entire little library (that added heading navigation and imenu support to shr) from my configuration! I *love* deleting code from my Emacs configuration.]
New post on M-x apropos Emacs! May I recommend eww for Emacs's innovative UI?
https://www.matem.unam.mx/~omar/apropos-emacs.html#may-i-recommend-eww-for-emacs-innovative-ui
Please consider this as entry for the May edition of the #Emacs Carnival, @sacha.
A master's student of mine wanted to go to a workshop in Canada and applied for a visa. He was rejected for reasons that sound insane to me: they said he didn't have sufficient family ties outside of Canada (his entire family lives in Mexico!); and they said it wasn't clear that he would benefit from attending the workshop (which I consider to be an academic decision that this embassy employee was not qualified to make!).
I've seen many, many students be rejected for American visas, but I think this is the first time I've seen this happen with a Canadian visa.
