Pinned Tweet
Check out our new open science tool that brings universal learning right to your browser! It uses algorithmic information theory to discover patterns in data - no machine learning expertise required.
openscienceresearchpark.com/
🧵(1/3)
Earlier this morning, @unclebobmartin posted a video from his Cybertruck showing a couple of women vandalizing it by writing "N*zi" on it.
I thought to repost it since I'd like to discourage that kind of behavior.
Within hours, it was taken down.
Looks like a great tool to integrate with agents for VLSI.
My IFIP WG2.11 talk is done and I can finally relax and enjoy the other talks. I showed several examples of deriving efficient circuits: a carry-lookahead adder, CRC using an LFSR implementation (and a software table version), and four kinds of sorters. All the code and proofs can be found at github.com/satnam6502/bird-m…
Rudi Cilibrasi retweeted
My dear friends, I’d like to show you my latest toy project
Introducing Handwired: turn your regular JavaScript code into a neural net!
- no training
- exact outputs vs the JS program
- fully differentiable
All runnable in a browser, without GPU
I wonder who has heard of the "Formula of Concord" ?
en.wikipedia.org/wiki/Formul…
Rudi Cilibrasi retweeted
This rule, against rewriting code in a different language, has become obsolete. It's important to understand why.
Historically we rewrote code by hand, which inevitably introduced bugs. And we weren't very rigorous about writing the regression tests that should have smoked out the bugs after translation.
Today, we have robot friends to do our code rewriting for us. And, if we have any trace of a freaking clue, we have already written regression tests with excellent coverage.
Under these circumstances, an automated rewrite in order to collect performance gains or memory safety can be a good thing to do.
Generally speaking, these days, if I have to do major surgery on one of my projects in Python or C, I will enlist a robot friend to lift the code to either Go or Rust. This has never failed to be a functional improvement.
Replying to @esrtweet
One thing I learned EARLY in my career: if it ain't broke, don't fix it, and FOR GODS SAKE DON'T REWRITE IT.
Rust breaks both of those.
I Love Lucy 2.0!
youtube.com/watch?v=LnXTPB98…
First hour since I woke up I already read:
1. An excellent X thread on "soundness" comparing practical techniques with various theoretical boundaries and interpretations of the word.
2. A solid review of Godel/Tarski (first one today, had one yesterday).
AI made math cool.
Replying to @renegadesilicon
True. Formal systems are the point.
But “formal” doesn’t have to mean the specific game Gödel and Tarski were playing.
You can still have precise, checkable systems that drop one or more of their assumptions:
infinitary rules of proof
built-in external semantics (truth as a relation to models, not an internal predicate)
interactive / open systems that formally allow queries outside themselves
other logics (intuitionistic, linear, paraconsistent, etc.)
The limits they proved hold inside that particular closed, recursively presented, finitary setup.
They don’t automatically transfer to every other formal game.
GenX here and I’m going to demonstrate something boomers can’t:
I know the generations born after mine are dealing with a more messed up world than the one I grew up in. I know everything is more expensive.
I know it’s harder and depressing and I know I had it easier.
Rudi Cilibrasi retweeted
Robots will need to download skills to perform rare tasks, and for that, they'll need your credit card.
I have never before seen this theoretical direction addressed before but I've long held this intuition.
Thanks for the heads-up @SMT_Solvers ! Amazing work, @PassRafael et al!
Why care about Kolmogorov complexity?
LLMs = monotone compression
(and whether “monotone” is necessary hinges on whether one-way functions exist)
arxiv.org/pdf/2610.11031
I thought Princeton denied her tenure decades ago so she went to Emory.
Congratulations to Anne Carson on receiving the 2026 #NobelPrize in Literature.
Carson taught at Princeton University in the 1980s and was recently on campus to deliver the 2025 Fagles Lecture: bit.ly/3TteJZj
Readers added context they thought people might want to know
Anne Carson taught at Princeton in the 1980s as an assistant professor but was denied tenure and left in 1987.
en.wikipedia.org/wiki/Anne_Cars…
thenation.com/article/cultur…
Information Theory for Neural Net modular memory?
Instant follow!
LLMs surprisingly store facts at the information-theoretically optimal rate. They do so in their MLP weights. How is this possible?
To answer this question, in our #COLM2026 paper, we show how MLPs can be constructed, without training, to store facts at the information theory optimal rate while remaining usable within transformer blocks for factual recall!
Rudi Cilibrasi retweeted
Some initial thoughts on OpenAI's incredible paper drop (UGC, ω ≤ 9/4, L = BPL) and some earlier AI-assisted work on my open problems.
blog.computationalcomplexity…
And a website to make it a little easier to browse OpenAI's TCS-related results
lance.fortnow.com/openai-mat…




























