rank decomposition
rank-decompositon.bsky.social
rank decomposition
@rank-decompositon.bsky.social
researcher, curator, forever student
discord @ rank.dim | email @ request
Reposted by rank decomposition
Fairy driven development
November 27, 2025 at 5:16 AM
Reposted by rank decomposition
Noether didn't just revolutionize physics — she introduced the kind of structural thinking that is now part of the water mathematicians and programmers swim in. Haskellers who point at various operations and say "this is a monad" are following in the footsteps of Noether
March 12, 2025 at 5:59 PM
Reposted by rank decomposition
I picked crafting interpreters + esolang 2 days back and @dtemkin.bsky.social upcoming new book popped on my feed. Nature ( aka algo gods) giving signs I am on right path.
Can You Code With Pictures? Esoteric Languages Challenge How We Code
Esoteric languages spark creativity by challenging conventional coding. Daniel Temkin is writing codes to confront the lack of creativity in AI code.
spectrum.ieee.org
September 16, 2025 at 12:51 PM
still standing
next worst thing would be a 1k+ stars 'Awesome'-list github repo
September 15, 2025 at 8:27 AM
Reposted by rank decomposition
the real reason it's a poor model is that modern LLM chatbots are not actually modeling next-token prediction

the pretraining objective is "predict the next token", but the post-training objective is closer to "create a response that is correct, properly formatted, and in line with style+safety"
"It just predicts the next token based on data in the training set" is a poor abstraction of LLMs because it makes poor predictions

For example, we might predict that it's very difficult to have an LLM emit the phrase "the quick brown fox jumps over the lazy cvnpmnzq", but it's trivial
September 4, 2025 at 9:41 PM
Reposted by rank decomposition
the more you know
For a moment I assumed that neuro-sama must has been an abbreviation from neuro-Sam Altman, but they're actually an LLM-powered vtuber huh
September 9, 2025 at 6:50 PM
Reposted by rank decomposition
LLMs as rubber-duckies
September 7, 2025 at 3:28 PM
Reposted by rank decomposition
machine demonology research
From my intro lecture to the undergrads
September 7, 2025 at 3:21 PM
> Half of AI agents are blind; the other half are deaf.
Current AI agents are perceptually limited. To illustrate how, I created a site that only some AI agents can see.

> New blog post: "The Browser Sensorium"

gracekind.net/blog/browser...
The Browser Sensorium • Grace Kind
Half of AI agents are blind; the other half are deaf.
gracekind.net
September 4, 2025 at 12:04 AM
August 29, 2025 at 2:47 PM
read this as 'autoregressive'
im not cooked but deep fried
the competitive advantage of aggressive kindness
August 29, 2025 at 2:44 PM
to not to be a content creator
August 29, 2025 at 2:41 PM
Reposted by rank decomposition
general question, was mathematics invented in the right order
How is it that vectors and matrices weren't a thing until the 19th century but analytical mechanics and functional analysis were already things in the 18th century?????
August 22, 2025 at 5:37 PM
Reposted by rank decomposition
math.univ-lyon1.fr
July 17, 2025 at 10:28 PM
back here for my weekly visit
July 17, 2025 at 10:11 PM
realized I dont visit this site as often
July 11, 2025 at 9:35 PM
saw a website embedding bsky posts as comments
July 8, 2025 at 6:47 AM
why does bluesky feel 'static'
July 7, 2025 at 9:12 AM
looks like its time to convert old laptop into a nix machine
July 6, 2025 at 1:00 PM
some fear mediocrity I specifically fear ending up an educator
July 5, 2025 at 11:43 AM
what is 'citable' exactly ?
July 4, 2025 at 7:09 PM
I am here to play the long game [this time]
July 4, 2025 at 11:10 AM
I have accepted it my fate to get doxxed
July 3, 2025 at 5:12 PM
one day one of these domain registrar will have a data leak, and we will all be naked
July 3, 2025 at 3:39 PM
I am doing this I am serious about it just you wait you need to wait a little
I am going to play with bluesky API
July 3, 2025 at 2:53 PM