Compression and LLMs are trying to solve the exact same problem: predicting what comes next. @anniesexton.com's interactive essay walks us through the basics of compression to reveal its surprising overlap with every language model you've ever used.

Maybe @masnick.com is right and coding with LLMs will lead to a Geocities-que renaissance of the weird, the peculiar, the quirky on the web

Anyone have recommendations for user-friendly or even semi-user friendly GPU platforms where I could spin up a vLLM server without breaking the bank? Tried Modal, but my container image builds are breaking :(.

Logged off of Twitter for prob the last time. It's not a place I've wanted to be for a long time. Even if you ignore all the other bullshit (and you shouldn't), the public tech agora of the 2010-2020 era is gone now. I miss the conversations. I hope they will continue somewhere in some form.

In the last years, I spent some of my scarce free time writing up what I had learnt starting and running two companies, zynamics and optimyze The article is still work-in-progress, but I thought it's best to publish it now and iterate in public: thomasdullien.github.io/guides/entre...

btw, slightly offtopic, but are there any simulations/calculations on whether algae could be used for carbon removal at an industrial scale?

Katie Mack@astrokatie.com · 2mo ago

Algae are prolific producers of oxygen (and consumers of carbon dioxide), en masse producing a huge proportion of the oxygen we breathe, which means that the greening of the Reflecting Pool might be the most pro-climate thing Trump has ever done.

The fact that Claude allows me to be fairly productive (on small matters) in a codebase written in a language (Rust) that I have very little experience with is a small miracle. The downside is that I can only be productive in this codebase with Claude. If Claude is not working, neither can I.

I sound like a broken record, but I really miss having an online place to discuss tech stuff. In spite of all its shortcomings, tech twitter c. 2010-2020 was really the peak.

mood rn.

the hungry observatory

Little did past me - furiously writing Twitter quitpieces for my blog- know that one day, I would come to sorely miss the place known as Tech Twitter.

ok, I'm intrigued

NeurIPS Conference@neuripsconf.bsky.social · 3mo ago

The call for the NeurIPS 2026 Creative AI Track is out! In its fourth year, NeurIPS 2026 Creative AI Track invites research papers and artworks that explore emerging applications, methods, and critiques of artificial intelligence and machine learning in art, design, and creative practice.

Hey, I wrote a thing about AI in astrophysics ergosphere.blog/posts/the-ma...

This is so, so well-articulated.

Minas Karamanis@minaskar.bsky.social · 5mo ago

Hey, I wrote a thing about AI in astrophysics ergosphere.blog/posts/the-ma...

Really miss workplace environments where there is low ego and hierarchy and everyone is just learning from everyone regardless of experience or tech stack.

I’ll bite. A machine that: 1. Experiences the world with us 2. Shares some truth-interest in the world (intellectual integrity & curiosity) 4. Forms ideas rather than barfs up chunks of algorithmically masticated text 5. Created to complement rather than devalue human intelligence and creativity

This is how I learnt about GPUs without basically any background knowledge.

dan@danabra.mov · 10mo ago

this is the way to learn with ai. you can start anywhere and backfill. you have to remain curious and careful and not just passively eat up plausible explanations. but if you put in the effort and the model is good, it’s powerful

modern LLM inference engines like vLLM & SGlang are becoming tough to dive into. to learn how these inference engines work, nano-vllm is a fantastic educational project—complete Page Attention & LLM scheduler in <1k loc.🤯 flaneur2020.github.io/posts/2025-1...

A Walkthrough of nano-vllm | Flaneur2020

Recently, I&rsquo;ve been delving into the architecture of production-grade inference engines. While projects like vLLM and SGLang are crazy sophisticated, …

flaneur2020.github.io

TIL that junk journaling is a thing!

Microblogging like it's 2010s with vibes.

Me after hearing that Python 3.14 has removed the GIL: "ah, finally increased throughput of pulling data from Big Query into my Jupyter notebooks." Also me: "ah, a new footgun to add to my repertoire"

One of the design issues with Jupyter notebooks when it comes to heavy ML workloads is that the notebook server runs by default on the same machine as the kernel that executes the code.

NVIDIA seems to invest a lot of engineering effort into making higher level libraries for writing efficient GPU code and yet everyone is flexing by rolling out their own CUDA kernels.

A speech about what drives me, how science and open source are bitter victories, unable to make improve the world if society does not embrace them for the better: gael-varoquaux.info/personnal/a-...

A national recognition; but science and open source are bitter victories

I have recently been awarded France’s national order of merit, for my career, in science, in open source, and around AI. The speech that I gave carries messages important to me (French below;...

gael-varoquaux.info

I was looking for a solution to "migrate a container that is close to OOM" onto another node and found CRIU. Still a bit unclear if it supported on Google's GKE or not.

Container image experts - is it possible to manually create a new layer by manipulation the files in the tar archive you get after running docker image save?

One of the major design flaws of many notebook environments like Jupyter is that the kernel that does computations is not separate from the machine that runs the notebook server itself.

When folks say they are going to build AGI - what exactly does that look like?

I really miss London.

I am very much a beginner in this space but napkin math for systems problems is quite a lot of fun!