"AI data center" moratoriums are not the way. They risk pushing burdens to places with less regulation, and driving up the price of compute, which will worsen inequality. jacobin.com/2026/04/ai-d...

Uh based alert? jacobin.com/2026/04/ai-d...

Benjamin Franklin wrote down his dumb jokes and random thoughts and every year he collected them and took them TO A PRINTING PRESS and distributed them. He did this for decades. That’s how much of a born poster he was.

I don't actually think this, but every time I see one of these I have to convince myself it's not CGI. I think it's because they're made of plastic and I've watched too many sci-fi movies so wires are getting crossed

seattle today

2%, "Below random"

The complex plane colored by the argument of a polynomial. Each root drifts along a tiny orbit and the color vortices follow: every swirl is a zero. Made with #python #numpy #matplotlib in a @marimo.io notebook.

presumably then everyone frantically tries to convince everyone to use their preferred strategy for saving everyone, including with precommitment devices like "everybody we know promises to vote blue except Bad Sam, who will vote red, and in the case red wins, hunt everyone who survives for sport"

no more thought experiments

the purpose of a system is what it thinks

You can just learn things!

Anyone else starting to accumulate Claude Code windows the same way as with browser tabs and open PDFs?

Still can't really say anything about quantum consciousness, but I think we're not too far off from being able to conclude that generic intelligence is just a classical phenomenon, empirically

Increased judgmentalness of ridiculous user requests as signal of intelligence Need a benchmark

My new thing is, cd to an ancient project directory I vaguely remember but have no idea what it does or how to run it open Claude Code in the before times, that project would have been lost to time and memory

I think that academia has not absorbed the fact that AI agents are now good enough to independently reconstruct complex papers without access to code or the papers themselves; just the methods & data. They aren’t perfect but the errors are often in the human paper, not the AI making a mistake.

Factorio has sold over 8 million copies. The people yearn for automation so hard they do it recreationally.

Gibson actually calls this out in the introduction to the 2004 edition of Neuromancer: "a whopping anachronism, right at the very start of my career in the imaginary future." It's a thoughtful reflection on the book's 20th anniversary

> User's collaboration style: technical, prefers honest scope re-assessments over forced optimism; says things like "never give up never surrender" and "time is irrelevant only progress" when wanting me to push on

People really, really loved the idea that ordinary messy humanness was itself a scarce resource that no machine could ever replicate. We've loved that story a long time. It's the moral of a bunch of old Trek episodes. "Synthetic data can be better" is the worst sort of heresy.

it is amazing how often people with zero knowledge about LLMs are confidently wrong about LLMs. like it isn't hard to go look up whether model collapse is happening (it isn't) and yet it just seems intuitively obvious so they don't fact check. yes I am subtweeting but it's a post on a different site

I was briefly obsessed with Venice as a kid because I thought it was like an American suburb except everyone drove boats instead of cars Remembered this when I visited for the first time recently and realized that of course you just walk everywhere, because it's a city in Europe

Man I have a really hard time thinking about the anthropic principle now without getting tripped up on the word to the point where I think I've misremembered it

LLMs working ok with fp4 gives me cosmic horror jeebies

How does this work, anyways? I was worried navigating things like social dynamics between drivers and near construction would be an unanticipated challenge and certainly requires complex reasoning outside the simplest world model for a car, but it sounds like they're doing okay

Shitposts too cheap to meter

does anyone else find compaction kind of a sad moment

I had a range of AI models "build me a procedurally generated 3D simulation showing the evolution of a harbor town from 3000 BCE to 3000 AD" in one prompt. You can play the full gallery here: hg-20f7d1a3ce.netlify.app Or read my write up about GPT-5.5 here: www.oneusefulthing.org/p/sign-of-th...

So how's the new DeepSeek? Looking for user reports

Imagine that at 2:02 am on Apr 10, 2011, a strange woman bangs on your door claiming to be your future self. Wild eyes darting, hypervigilant. Hair like a bowerbird's nest. Was she dressed by a tornado? "You don't UNDERSTAND," she says. "I need to STOP it!!!" ...stop what? "The SINGULARITY!!!!!"

Let Claude sleep!

gpt5.5's Weird Little Guy behaviors are really something

Open models have one month left to properly catch up with Opus 4.5 / GPT 5.2 to maintain the usual "6 months behind" heuristic

It'd be nice if Codex didn't need npm to install, Claude Code's simple curl | bash thing really lowers the barrier to just trivially sticking it wherever

"people in the past thought brains were steam engines" has this identical form. yes but every single correct theory was also preceded by past lineage of incorrect theories, so if we couldn't read the theories that would provide maybe a tiny jot of ratio evidence but since we can it's ~worthless

"All energy is solar except nuclear" is incorrect; uranium is produced in supernovae, so it's still solar, just a different sun Only when we get fusion of light elements online will we be freed from this tyranny

Has anyone spotted a spontaneous "ope" or "welp" yet? Midwestern Claude

Maybe pasta sauce is just Italian chili

Minority Report -> Hot Tub Crime Machine

here come old laptop he come bootin up slowly he got broken touchpad he need one more update he got no space on this pc one thing i can tell you its got no warranty

one of the recurring characteristics of this whole thing is that it is moving crazy fast. people tend to read that as “it is moving so fast that only insiders understand what’s happening” but I think it is “it is moving so fast nobody understands what’s happening”

two short business weeks later, your commission sir @abeliansoup.bsky.social ; @codetard.bsky.social

LLM making a joke based on its own "lived experiences"?

The jagged frontier

This is exactly the same cause as every valid LLM excitement about societal effects

yes! yes! it’s all just computerization www.dwarkesh.com/p/ada-palmer

Every system that was regulated, either explicitly or implicitly, by the fact that they were effortful for humans (letters of recommendation, government filings, essays, or, as this paper finds, lawsuits) will break under a wave of AI.