Nicholas Dwork

“Holey Cow” - a children’s book. Written by me. Illustrated by my mom (@AnaDwork). a.co/d/czL7N7f

@atmoio LLMs get a simple prototype up and running sometimes. As the project becomes complicated, when it no longer works and AI can’t fix it, I find that I’m unable to wade through the mass of code created and address any issues. Like any devil’s bargain, it seems we lose.

I was a 10x engineer. Now I'm useless.

Transformers are Bayesian Networks arxiv.org/abs/2603.17063

This is one of the truest and funniest things I've ever seen.

@SinkingStarGame Why?

Over a thousand puzzle fans have already joined our Order of the Sinking Star Discord 💛 🫵 You should definitely do so too: discord.gg/PcCxqvfeKK See you there!

This is a simple alternative to low-dimensional embedding methods such as tSNE, UMAP, and PCA. It trains an autoencoder to match the distances of the reference space. The results quite good. 🔗arxiv.org/abs/2603.16568

Jeremy's (@jeremyphoward) and Terence's (@the_antlr_guy) backprop guide is sooo well written that 3 years later as I'm reviewing the guide, everything is just coming back so fluidly. 11/10 would read again explained.ai/matrix-calculu…

@Mormor0203 @BasedMikeLee @johnthune @LeaderJohnThune Cloture requires 60 votes. Here's hoping @LeaderJohnThune doesn't have them!

@BasedMikeLee What is the plan if @johnthune @LeaderJohnThune files cloture? We need to have one!!!

No Senate rule changes required

@BasedMikeLee Can you give us your assessment @BasedMikeLee ? What are you expecting in the next few hours?

@FreyaHolmer Will it be recorded and made available?

I'll be speaking at GAME 2026 next week in Kortrijk, Belgium! bivector.net/game2026.html

@FreyaHolmer Love it.

been feeling kinda stressed lately so I made a little prototype is this anything

> If you’re reading my article about probability, prediction markets and tools like MiroFish > this 1 hour Stanford lecture “Bayesian Networks and Learning” is worth watching > one of the best short introductions to how beliefs update with new information > highly recommend

After 18 months of hard work by Tomas and Zhen, we cooked it! 🚀 Thanks to all friends who give constructive feedback! Deep Learning 2.0, Rethinking every fundamental cornerstone of Modern Foundation Models. It's just the beginning, Hyped! 🚀 github.com/FlashSampling/…

Finally finished! If you're interested in an overview of recent methods in reinforcement learning for reasoning LLMs, check out this blog post: aweers.de/blog/2026/rl-f… It summarizes ten methods, tries to highlight differences and trends, and has a collection of open problems

I (finally) put together a new LLM Architecture Gallery that collects the architecture figures all in one place! sebastianraschka.com/llm-architectu…

My view today

An absolute banger of a paper. “A Gentle Introduction to Matrix Calculus” by econometrics legend Jan Magnus — one of the clearest explanations of matrix derivatives ever written. If you work in econometrics, machine learning, statistics, or optimisation, this paper is pure gold.

@JohnStossel I taught my daughter to read with a book using phonics. It worked very well. She's not in kindergarten yet, and already reading simple chapter books. a.co/d/02LS5KZs

“My child can’t read!” That’s become a common complaint from parents. Why? It might be because kids are distracted by social media and video games. But I think it’s also because reading instruction became lazy and political. johnstossel.com/the-reading-wa…

@Danukeru @mathelirium This is an algorithm to find k dominant Fourier values in a short amount of time. Compressed sensing recovers a sparse vector from a small number of measurements. Commonly, in compressed sensing, it's not the Fourier measurements that are sparse.

@mathelirium How is this different from compressed sensing?

This video presents a recent breakthrough called the Sparse Fourier Transform (SFT). This algorithm yields an exponential speed-up over the celebrated Fast Fourier Transform (FFT) when asked to extract a small number of dominant Fourier coefficients.

@modonnell Homeownership is especially tough given the ever increasing amounts we have to pay in property taxes.

Colorado consumers each held, on average, $72,210 in outstanding mortgages on 12/31/25, the highest average of any other state. It is worth noting that Coloradans did a much better job than residents of any other state in keeping mortgages obligations current through the worst of the 2007-2009 recession.