me: if z is the length of a slice and a is the area of the pie, then pi(zz)=a
PhD advisor: this is what you’ve been working on for three and a half years?
15.03.2025 14:44
👍 1330
🔁 257
💬 19
📌 4
me: if z is the length of a slice and a is the area of the pie, then pi(zz)=a
PhD advisor: this is what you’ve been working on for three and a half years?
A nice article by Nicholas Carlini on writing great papers.
A lot of the advice you can find elsewhere but the sections on collaborations and conclusions both had new insights for me.
It also helps that I've really liked the papers of his that I've read
nicholas.carlini.com/writing/2026...
Same here!
It's a shame it's actually showing a much less fun kind of uncertainty
It also explains why the prequel trilogy was so controversial. All the CGI meant that, subconsciously, everyone was asking themselves "but where are all the Muppets?!"
But what if that's because Star Wars is already a Muppet movie?
A Muppet movie is a bunch of puppets and people in costumes interacting with straight-faced human characters, providing a mix of earnest advice and comic relief.
That sounds a lot like the original trilogy
A simple command line tool combining a crossword solver and anagram finder.
A lot of online crossword solver tools felt unnecessary cumbersome to me, so I've tried to make this as lightweight and simple to use as possible
github.com/EchoStatemen...
Oh, so this is what people mean when they say "do it for the plot"
You know that thing where the farther something is, the smaller it looks? That doesn’t quite work in an expanding universe. It’s true for a while but after a certain point, more distant things (of the same size) start to look BIGGER because their light has been traveling since the cosmos was small 🙃
This post was inspired by a video by @tomvii.bsky.social, where he uses paraphrasing by large language models as part of an entirely new typesetting system based around his own programming language
I highly recommend checking it out:
www.youtube.com/watch?v=Y65F...
Image showing the transformation of the intro paragraph to the Hot Fuzz Wikipedia page into perfectly justified monospace text
Can we get LLMs to generate text that is perfectly typeset down to the very last character?
In a new blog post, I show how using a constrained beam search, LLMs can efficiently generate perfect blocks of monospace text
echostatements.net/posts/2026/0...
The library is open source, so feel free to play with it:
github.com/EchoStatemen...
I've added the ability to create Delaunay triangulations to my Voronoi diagram generation Python library.
The image generation takes a few seconds but I'm pleased with how easy it is to make pretty and varied outputs with it
TIL if you do Miller-Rabin with the first 12 prime bases, you have a fast, exact primarily test for all 64-bit integers.
Cf. oeis.org/A014233 which gives the upper bound on the two-sided correctness of Miller-Rabin given that you tested on the first k bases.
The lego logo
A lego forklift
And an honourable mention:
Lego: Lego
Lego supposedly are the world's largest manufacturer of tyres (at least according to the Guinness*). Sure, tyres aren't their main focus, but that doesn't seem so unusual for a tyre company!
* While we're at it: yes, Guinness the brewery/word record authority
The good year logo
The Goodyear blimp
Goodyear: Blimps
Unbelievably left-field. In the 20th century, they at least built the blimps themselves, but now their sideline seems to just be being the world's most famous owner of a fleet of airships
The Michelin logo
A fancy looking meal
Michelin: Restaurant reviews
This clearly got out of hand for them. It started as a more general guide for motorists, (even that is a slightly eccentric thing for a tyre company) but getting your tyres from the world's foremost authority on restaurant quality is odd
The Dunlop logo
A Dunlop tennis ball
Dunlop: Sports equipment
If you're working with rubber anyway, I can see how you get into making tennis balls and how that might spiral into sporting goods more generally
Fair enough
Tyre companies and their increasingly unhinged side projects, a thread:
Suddenly, the explanation in Solo for how Han Solo got his full name doesn't seem quite so far-fetched
New tragedy of the commons just dropped
www.bbc.com/news/article...
Hahaha, I want to get that on a t-shirt
It was definitely one of the things that got me hooked on bias-variance decompositions, and more generally on information geometry... Even if I'm not especially active in either any more
I guess you never know what's going to resonate with people!
I had always wondered why you hadn't done more with it. FWIW, we didn't find Buja et al's result until long after yours, so having it there was a huge boon to us
I'm really glad to see this on arXiv after all these years!
As David points out, he wasn't the first to discover this result, though he certainly deserves a great deal of credit both for its rediscovery and for (apparently somewhat accidentally) popularising it with a punchy paper/note containing clear and succinct proofs
This result was one of the foundational ideas built upon by what has since become by far my most cited paper, so I'm really happy to see a version on arXiv where it can hopefully be preserved for posterity
You could say that having chosen a reference distribution, the KL divergence tells you how far away another distribution is in the particular way that you care most about
It's vaguely distance flavoured, but in a way that we don't (as far as I know) have a general word to describe
I disagree slightly that it's not just semantics: when deciding how to generalise the notion of distance, we get to decide which attributes have to be retained in order to retain the "distance" label
Requiring symmetry is perhaps the most reasonable choice but it's not the only possible choice
I think there's a divide between those who use "distance" synonymously with "metric" and those who use it for any function with a more general sense of dissimilarity-ness.
I wonder if the former group want to say KL isn't a distance whenever possible to make their definition the more prevalent one
