Slava Pestov

Programming languages, finitely-presented monoids, horses

Public Key

npub1wlz26ycfd7gr7jhqshvl0jqsuj2pg2nydcrqhpn95l50teznvwgq93axm0

NIP-05 Address

[email protected]

Profile Code

nprofile1qqs8039dzvyklyplftsgtk0heqgwf9q59fjxupstsej60684u3fk8yqpz4mhxue69uhhyetvv9ujuerfw36x7tnsw43qwvjuak

Publishing to

relay.ditto.pub

Author Public Key

npub1wlz26ycfd7gr7jhqshvl0jqsuj2pg2nydcrqhpn95l50teznvwgq93axm0

Show more details

Last Notes

2026-11-02 14:46:25 GMT

Random access memory is so expensive now programmers are going back to Turing machines

2026-10-29 00:45:56 GMT

I love it how this guy turned his mid-level tech career into a whole influencer brand.
“Did I mention, I’m a staff engineer?”
https://media.mathstodon.xyz/media_attachments/files/117/521/427/174/434/287/original/ee361ef7f2bdd564.jpeg
https://media.mathstodon.xyz/media_attachments/files/117/521/428/055/954/164/original/eac94103cb98ead9.jpeg
https://media.mathstodon.xyz/media_attachments/files/117/521/428/092/019/762/original/0153d22fbc7f3526.jpeg

2026-10-28 12:06:48 GMT

I’m torn on this. On one hand. TeX *itself* is a beautifully crafted and well-documented piece of software.
But, once you layer LaTeX and a pile of random CTAN packages on top, some problems become apparent:
- poor error messages
- slow compile times
- packages that must be loaded in a certain order, or can’t be used together at all in one document, etc
I’ve never used Typst so I cannot compare, but I think TeX ought to be replaced at some point. It has served us well, but at this point it’s more than 40 years old. Which in itself makes it an incredible achievement of course!
#nevent1q…f249

2026-10-28 02:52:37 GMT

How about instead of "Super Intelligence" we call it "American Intelligence"

2026-10-26 14:31:33 GMT

Gforth 1.0 has been released
https://gforth.org/manual/1.0/NEWS.html

2026-10-26 00:07:12 GMT

Good news, Bolsonaro’s failson is cooked

2026-10-24 02:06:18 GMT

It was always my destiny to learn this stuff, turns out I even named a programming language after it without realizing
https://media.mathstodon.xyz/media_attachments/files/117/493/428/144/233/561/original/1af570b53a72fc2e.jpeg

2026-10-23 22:48:00 GMT

There are two languages where I'm still constantly discovering new features years later, C++ and Markdown
#nevent1q…eq4l

2026-10-23 12:24:17 GMT

Slop code is bad enough but LLM-generated PR descriptions are a whole new level of brainrot

2026-10-22 00:48:56 GMT

But I *like* Perl
#nevent1q…z72t

2026-10-13 23:50:02 GMT

Sam Altman is upset someone left a recording device in his conference room. I guess he forgot to explicitly opt out of data collection for marketing and training purposes

2026-10-11 02:42:40 GMT

Nobody had the heart to say “okay Ronald, that actually sounds pretty weird”
https://media.mathstodon.xyz/media_attachments/files/117/419/968/168/202/658/original/4c8244c1aa635127.jpeg
https://media.mathstodon.xyz/media_attachments/files/117/419/968/107/478/794/original/cb17c2b5c9d5341c.jpeg

2026-10-09 18:59:42 GMT

A notable feature of Common Lisp is that it has complex number arithmetic, complete with all elementary functions, as part of the standard library. I do wonder though, how many Common Lisp programmers have ever needed to use this?

2026-10-07 15:39:38 GMT

The noses came over to say hi
https://media.mathstodon.xyz/media_attachments/files/117/400/375/331/316/751/original/01752f9b89a1cee0.jpeg
https://media.mathstodon.xyz/media_attachments/files/117/400/375/376/609/190/original/ca05d51bdf16ce35.jpeg

2026-10-07 02:01:56 GMT

Because of course the Tandy Color Computer had a Forth implementation named ColorForth, completely unrelated to Chuck Moore’s ColorForth https://forum.vcfed.org/index.php?threads/color-forth-available-for-coco-with-decb.46819/

2026-10-03 20:38:08 GMT

Evolution of programming languages at Apple:
- HyperTalk
- AppleScript
- NewtonScript
- Dylan
*meteor hits the earth, somehow AppleScript survives*
- Objective-C
- Swift

2026-10-02 21:23:52 GMT

He’s like a plush toy
https://media.mathstodon.xyz/media_attachments/files/117/373/413/323/251/061/original/6083fb42bfc85fe7.jpeg

2026-08-15 20:22:07 GMT

- reply

@nprofile…0f5x there are some really nice horse trails up in the hills between Cupertino and Redwood City! Unfortunately you’d have to deal with people who have both VC-level wealth and are horse people which seems to make them completely insane

2026-08-15 19:40:25 GMT

This ridiculous guy
https://media.mathstodon.xyz/media_attachments/files/117/101/215/747/333/908/original/94b841edf9c3e525.jpeg
https://media.mathstodon.xyz/media_attachments/files/117/101/216/541/453/366/original/a7fa8aa5f22570f3.jpeg

2026-08-13 04:48:38 GMT

- reply

@nprofile…0f5x It's a great book nonetheless!
Also, congratulations, you've now written more Swift code than most of the compiler team. Or maybe they're writing it in secret, because Apple's OSS policy doesn't allow sharing code in public

2026-08-13 04:45:27 GMT

- reply

@nprofile…0f5x Looks nice!
IMO, immutable global variables declared via `let` are fine, and preferable to singletons, which feel like needless boilerplate to me.
Also, you've got a few "if (...)" with redundant parens in there. C programmer spotted :-)

2026-08-07 07:56:52 GMT

“Haskell attracts idealists who care about correctness and think carefully about the abstractions they build with. Hiring for a Haskell role has never been the problem that leaders think it will be. If anything, candidates came to us for the Haskell itself, even when they had no particular interest in supply chain.”
#nevent1q…tkze

2026-07-10 15:01:04 GMT

The Rubik’s cube is a group until you apply a non-invertible transformation by smashing it with a hammer, at which point it becomes a monoid

2026-07-05 15:12:59 GMT

It’s annoying that programmers don’t generally try to write documentation, or see it as a task they should get good at.
Java’s “javadoc” generator kicked off this industry-wide trend of replacing all developer documentation with auto-generated API docs extracted from comments, which are invariably low-information “getFoo(): returns the current value of Foo”, often just written by the last person who touched the code, maybe without understanding it.
But even that was too much apparently, and now some programmers believe that with high-level languages and static type systems, code magically becomes “self-documenting.” The compiler prevents you from doing something invalid, and here is a list of the valid combinations. Oh, and here’s an “agent skill” file. Good luck!

2026-07-03 13:21:09 GMT

If you can solve the word problem in this monoid, you’re following too close
https://media.mathstodon.xyz/media_attachments/files/116/856/246/046/499/476/original/96e2c08156f42147.jpeg

2026-07-01 15:30:39 GMT

Smalltalk-80, HyperCard, and Visual Basic 6 were far more compelling "future of programming" attempts than anything available today

2026-06-29 16:16:56 GMT

This was a big piece of news from the other day. Finally I can start daydreaming about rewriting the compiler’s type substitution and requirement machine in Swift. There are several things that could benefit from an overhaul there, and it would be nice if instead of presenting pseudocode with a C++ source reference at the end of each chapter, the generics book could just show reasonable-looking code snippets from the implementation inline https://forums.swift.org/t/dropping-the-requirement-for-c-only-bootstrapping/87739

2026-06-26 18:28:19 GMT

Perl 5 LLM policy
https://perl.topicbox.com/groups/perl-core/Tded84d01cb16966e-M4f2b94e0c07f9d7cb8e1e76b

2026-06-23 15:57:58 GMT

Open source projects will tell you they have no choice but to allow slop, because doing anything else is a form of gatekeeping that will exclude too many potential contributors.
Nothing could be further than the truth, though. When the review queue is a non-stop machine gun of giant, bogus drive-by patches, the good contributions will inevitably fall through the cracks.
Eventually, the contributors you actually want—those who care about the long-term health of the code base, building institutional knowledge, improving their craft, mentorship—will simply leave, because they will realize their limited time and energy is better spent elsewhere.
A blanket ban on LLM contributions is not impractical, idealistic, or too radical—it's just common sense.

2026-06-23 13:03:58 GMT

“Avoid premature optimization” basically means, don’t make your design overly complicated in the name of performance unless you have good reason to believe it will be necessary.
If you interpret it to mean that you shouldn’t think about performance at design time because you can just profile the implementation later and fix all the hotspots, you’re going to have a bad time

2026-06-21 21:36:51 GMT

Dusted off (literally) the NeXTstation
https://media.mathstodon.xyz/media_attachments/files/116/790/245/358/749/034/original/4c82709049f9c3f3.jpeg

2026-06-21 14:25:05 GMT

https://media.mathstodon.xyz/media_attachments/files/116/788/553/029/067/480/original/33092790ba23a37b.jpeg
https://media.mathstodon.xyz/media_attachments/files/116/788/553/080/907/934/original/e7f936360543e53a.jpeg

2026-06-20 17:21:32 GMT

Vibe coders who say they’ve read every line of code that comes out of their slop machine are a bit like guys who think they’re good in bed because they’ve watched a lot of porn

2026-06-17 23:24:52 GMT

- reply

@nprofile…0f5x oh it was always bad there, the light turns red and then five more Teslas/Priuses turn left all tailgating each other

2026-06-02 23:49:13 GMT

Collatz semantic versioning: pick some integer to be your first version number. If it’s even, divide it by two to get the next version number, otherwise multiply by three and add one. When your software reaches version 1, proclaim it to be finished.

2026-05-27 15:23:08 GMT

- reply

@nprofile…0f5x jgroff sounds like something old that's written in Java, like jEdit. The Swift version could have a catchier name, like 'pomeranian'

2026-05-26 12:24:39 GMT

“We’re in the middle of a company-wide initiative to migrate everything to JIRA.”
“I thought you already used JIRA.”
“Yeah, but only for tracking bugs. We’re going all in now, and becoming a JIRA-first company.”
“What does that mean exactly?”
“Well, you see, we’re rewriting our app *in* JIRA.”
https://seriot.ch/computation/jira.html

2026-05-13 00:52:27 GMT

a + b = b + a
FINITELY
GENERATED
COMMUTATIVE
MONOIDS
2x + y = 0
y + 2z = 0
https://media.mathstodon.xyz/media_attachments/files/116/564/527/415/339/671/original/ec7a71e51d96e755.jpeg

2026-05-05 23:42:32 GMT

This article is about the software bug. For events occurring during the presidency of George W. Bush, see Presidency of George W. Bush.
https://en.wikipedia.org/wiki/Bush_hid_the_facts

2026-04-21 01:15:17 GMT

Oh nice, a new variant of SSA for directly encoding closures and higher-order functions just dropped https://arxiv.org/abs/2604.09961

2026-04-13 15:06:39 GMT

When I first started writing C++, I found myself making a lot of rookie mistakes, like returning a dangling reference to a local variable. These days though, I only make the kinds of mistakes that even trip up seasoned C++ experts, like returning a dangling reference to a local variable

2026-01-31 19:56:20 GMT

An Informal Introduction to Algol 68 https://softwarepreservation.computerhistory.org/ALGOL/book/Lindsey_van_der_Meulen-IItA68-Revised.pdf

2026-01-03 03:27:33 GMT

- reply

@nprofile…0f5x Four consecutive lines that end with hyphenated words?! Don’t tell Knuth!

2025-12-30 20:30:25 GMT

- reply

@nprofile…stct They’re just vintage!

2025-12-30 17:58:22 GMT

Math textbooks should come with their TeX source code

2025-12-28 04:27:51 GMT

I’ve always wanted an IBM mainframe! My brother-in-law gave me a LEGO z16 for Christmas. Check out the hot pluggable CPU
https://media.mathstodon.xyz/media_attachments/files/115/795/296/780/815/417/original/357d30128422b820.jpeg
https://media.mathstodon.xyz/media_attachments/files/115/795/296/788/151/950/original/979558a3315eeaf8.jpeg
https://media.mathstodon.xyz/media_attachments/files/115/795/296/787/958/353/original/b8ab660bdb684a27.jpeg
https://media.mathstodon.xyz/media_attachments/files/115/795/296/761/842/397/original/e8f2851078130703.jpeg

2025-12-24 23:50:31 GMT

If you have any remaining doubts that Perl is in fact the best programming language, ponder this: you can change array indexing to start from 1 (or any other integer, for that matter) simply by setting the magic variable $[
https://media.mathstodon.xyz/media_attachments/files/115/777/217/672/508/711/original/437642ca6f1f0d4e.jpeg

2025-12-24 20:53:51 GMT

- reply

@nprofile…08mx pro tip: it’s easier when the gifts are beer, then you just set a couple aside for yourself

2025-12-21 19:09:38 GMT

My three year-old calls full-size not-baby carrots “mama carrots”

2025-12-15 00:50:04 GMT

Reading a book about the J programming language. Check out the numeric literal syntax:
2e5 — 2 * 10^5, usual “scientific” notation for floats
2p5 — 2 * pi^5… what? So 1p1 is just pi
2r5 — 2/5, exact rational number
2j5 — 2+5i, complex number
(This is the least strange thing about J, by the way)

2025-12-12 14:59:42 GMT

Counter-Factual Typing for Debugging Type Errors https://dl.acm.org/doi/pdf/10.1145/2535838.2535863

2025-12-10 01:00:22 GMT

- reply

@nprofile…08mx wishing you a speedy and full recovery

2025-12-09 13:51:52 GMT

What a delight, a new release of my favorite concatenative language! https://re.factorcode.org/2025/12/factor-0-101-now-available.html

2025-11-30 15:45:22 GMT

The appearance of em dashes in someone’s writing is a telltale sign they used an LLM—or perhaps they managed to read as far as page 4 of the TeXbook. Sadly these days the former is far more likely
https://media.mathstodon.xyz/media_attachments/files/115/639/418/709/273/888/original/f654aa7955e03144.jpeg

2025-11-22 18:29:23 GMT

By the way, even if you have no intention of ever using it, the TikZ documentation is worth browsing just for the example pictures alone: https://tikz.dev/

2025-11-22 04:33:02 GMT

“The problem remained unsolved for a number of years, due partly to the difficulty of computing the huge numbers involved in the solution. The general solution was found in 1880 by Carl Ernst August Amthor, … Using logarithmic tables, he calculated the first digits of the smallest solution, showing that it is about 7.76×10206544 cattle, far more than could fit in the observable universe.“
https://en.wikipedia.org/wiki/Archimedes%27s_cattle_problem

2025-11-22 02:32:34 GMT

- reply

@nprofile…08mx if you don’t have a copy of the TeXbook yet, get one, you might enjoy the exercises

2025-11-19 21:30:06 GMT

Programmers have finally realized that a single function call can bring down an entire process, even in an allegedly “safe” language.
Just wait until they discover their compilers (especially the newfangled “systems languages with value types” that are so hot right now) have been lying to them this whole time by pretending that stack space is an endless resource, without providing any tools whatsoever for reasoning about how much of it they’re actually using

2025-11-12 16:50:24 GMT

This might be one of those things everyone else knows about already, but I just learned it now:
`git diff` accepts a --word-diff flag, which is handy for seeing what changed in the middle of a long line

2025-11-12 14:47:47 GMT

- reply

@nprofile…08mx Well, don’t leave us hanging! What’s the solution??!

2025-11-12 14:16:09 GMT

- reply

In case anyone is wondering, I didn’t use LLMs at any point while writing or editing any of the stuff I’ve posted on here about finitely-presented monoids or Swift generics. It’s still slop for sure, but it’s 100% organic hand-crafted slop

2025-11-11 18:00:09 GMT

I haven’t made as much progress on the generics book lately as I would have liked, but here is an update: https://github.com/swiftlang/swift/pull/85431
Earlier this year I spent some time debugging opaque result types, and while doing that I filled in the missing chapter which describes the feature in a semi-formal way.
I also reworked the substitution maps chapter with better explanations and examples. It was rather short, so I squeezed in two extra sections: the first is about subclassing generic classes, and the second is about SIL type lowering and re-abstraction thunks. Both are “practical applications” of substitution maps within the compiler, so to speak.

2025-11-09 02:28:43 GMT

- reply

@nprofile…08mx “Tomorrow’s integrals, today” is a good catchphrase for a computer algebra system ad

2025-11-08 14:31:21 GMT

The awk book is actually interesting to me because I’ve been using awk for literally decades now, but I always just copied snippets from Unix books and online sources without ever taking the time to learn it properly.
The first new thing I learned is that the action block that follows a pattern is optional! If omitted, it defaults to { print $0 }.
So for example ‘$3 == "Foo"’ is a valid awk program that prints all lines in the input file where the third field is “Foo”, ‘NR <= 10’ prints the first 10 lines of a file, and so on.

2025-10-30 17:38:12 GMT

- reply

@nprofile…stct @nprofile…60pe @nprofile…rrqe I know almost nothing about physics but I’ve enjoyed the tiny bit I have read, because the “non-rigorous” nature is actually a better fit for how I think about programming most of the time than math would be. When you have a large messy codebase you cannot hope to understand it exactly, but you can observe it and formulate theories that explain certain behaviors and then you refine those theories over time

2025-10-30 12:14:04 GMT

- reply

@nprofile…8eua Knuth 4A also spends a bit of time on the “majority” operation which takes three booleans and returns the one that appears at least twice

2025-10-10 20:59:26 GMT

- reply

@nprofile…nrqy Sounds like @nprofile…tfh0 forgot to _fixLifetime(cup)

2025-08-25 19:20:32 GMT

- reply

@nprofile…pd3k let c=ab and d=aab. Now a=aaaabb=aabaab=dd and b=aaabbb=ababab=ccc.
Also cd=abaab=aaabb=1, dc=aabab=aaabb=1, and aaabb=ddddddcccccc.
This means the original presentation is equivalent to <c, d | cd=1, dc=1, ddddddcccccc=1>, which is just
<c, d | cd=1, dc=1>.
What am I missing?

2025-08-23 20:10:19 GMT

“FLINT is a C library for doing number theory.
…
At its core, FLINT provides arithmetic in standard rings such as the integers, rationals, algebraic, real, complex and p-adic numbers, finite fields, and number fields. It also provides polynomials (univariate and multivariate), power series, and matrices.”
https://flintlib.org/

2025-08-23 17:31:46 GMT

The baby is taller than the pony now, but still only about half as wide. The Munch, munch, munch… sound is very relaxing
https://media.mathstodon.xyz/media_attachments/files/115/079/264/518/875/950/original/280c0ee5a5b9fa9d.jpeg
https://media.mathstodon.xyz/media_attachments/files/115/079/264/518/858/873/original/bd5934598df33c32.jpeg
https://media.mathstodon.xyz/media_attachments/files/115/079/264/556/686/349/original/6c40eff1097113d9.jpeg
https://media.mathstodon.xyz/media_attachments/files/115/079/264/578/108/595/original/f592d7be98b9db74.jpeg

2025-08-18 21:57:43 GMT

“Introducing our thinnest monoid yet.”
Here is a rough draft of a proof that <a,b|aaa=a,abba=bb> of length 10 does not have finite derivation type: https://factorcode.org/slava/aaaaabbabb.pdf
This also talks about my computational experiments with finitely-presented monoids.
This is a nice conclusion for a side project investigation (by “investigation” I actually mean an unhinged series of posts on this account) of two generator, two relation monoids of length 10.
I started looking at this 8 months ago over new year’s break. I thought it would be extremely funny if a computer search could uncover a two-relation monoid that cannot be presented by a finite complete rewriting system. I also wanted to do this with a minimum amount of code, so that I could distill the result into a compiler micro-benchmark, only a little bit more complex than your typical n-bodies simulation or whatever: https://github.com/swiftlang/swift/tree/main/benchmark/multi-source/Monoids
I didn’t really expect to get anywhere, but here we are! Enjoy.

2025-08-03 12:55:03 GMT

“If you’re not running a swarm of LLM coding agents, you’ll be left behind”
Bro I don’t even use code completion

2025-07-25 20:38:04 GMT

- reply

@nprofile…6cfc @nprofile…sast also, if for each generator there is an n such that a^-1 = a^n in the group, it seems that you can construct the group as a congruence on the monoid so again it has to be finite

2025-07-25 20:34:40 GMT

- reply

@nprofile…6cfc @nprofile…sast is $a^{-1} = a^n$ for some n though? Then $a^{-1} b= a^n b$ has finite order in the monoid and therefore the group

2025-07-25 15:43:14 GMT

I realized yesterday that I may have made a mistake in https://mathstodon.xyz/@slava/114767196106832634.
I assumed that if <A | R> is a presentation of a finite monoid, then the group with the same presentation is also finite. But I can’t show that this is true, and it might be false. Does anyone know?
In the meantime, we’re back to not knowing anything about the monoids <a, b | aaaa=1,aba=bbb> and
<a, b | aaaa=1,abbba=b>. They might still be finite, or at least I can’t say they’re not.

2025-07-13 00:51:26 GMT

- reply

@nprofile…nrqy Bravo. Continuous functions and uncountable sets are too much for me. Calling them “reals” is a stretch IMO