A lot of left critique is stupid and ignorant and deserved response but if your default posture is defending tech companies against posters you have fully lost the plot.

Gotta say I’ve never received a less enticing job offer than a cold email from your clawbot. Maybe just not a culture fit

I know I’m like 6 months late to thinking about this but it was actually pretty fucked up when a lot of you on here suggested the “clanker” people were as bad as racists. It is perhaps a sin to express vitriol in this way, but it is actually not comparable to white supremacy.

really annoying you can’t soft block people on here

38,55807° N, 76,08327° W if anyone in Maryland is interested

This resonates mariozechner.at/posts/2026-0...

After consideration my opinion about this and futurelock (same issue) is that tokio’s synchronization primitives are fundamentally buggy: they should not be using their internal semaphore to guarantee FIFO access

climate crisis activist Donald Trump will really be like 'you believe in voting? that pales in effectiveness to my strategy, firebombing oil and gas infrastructure' and then repeatedly firebomb oil and gas infrastructure

Really annoying to me personally that there’s no NonMaxU64 etc in Rust given that’s way easier niche to use if it’s an index into something. Of course it’s stalled on the maximally expressive implementation where you can have any niche range… YAGNI

Its kinda cool that computers are slow again now, in the way I imagine they must have been in the 80s

10 years ago I was very interested in critiques of cybernetics (eg wiki.fuckoffgoogle.de/images/8/8f/...) because I thought Hillary would obviously win we lived in a post historical technocracy. How far we have fallen.

The core concept of cybernetics is that we can, by iteratively feeding system output back into the system improve the steerage of the system toward our intended goals. It comes from the Greek for “pilot.”

the world does not need anymore videos of people holding a clip on mic in their fingers and giving their incoherent take on a news item they half understood

2026 is the year we be starting doing land acknowledgements for the braincoders yall

Backpropagation and gradient descent also have obvious limitations, and these very limitations made them impractical in those decades. Specifically they require data and compute that were not available. 1960s-1970s academia/industry were quite pragmatic actually

It’s actually a great time to care about the craft. I don’t know if people realize this but you can actually write whatever you want into the prompt box. You can use the AI to get the function just right instead of making it generate crap as fast as possible.

This was unfair. What’s actually remarkable about Java is that all of its basic affordances are well chosen, and then implemented with completely wrong formalisms. This makes it the perfect inverse of Haskell, which has immaculate formalisms for god awful affordances.

its so incredible the amount of information you can fit in 16 bytes

got some shocking news for people who post like this about information that has been written down

i love logging on to this app to piss myself off at the small minority of people on earth who 95% agree with me

So often “America is going to overplay its hand and become a rump state” basically seems to be the imaginary just punishment that stands in for the real horror, which is that America can ruin millions of people’s lives at the whim of a decaying moron and his Nazi friends and nothing happens.

it’s actually reprehensible to contract for the DoD in this war even if your technology isn’t responsible for targeting a girls school

Its such a funny affect that Claude has where I ask if code is right and it very confidently begins explaining why it isn’t until it realizes it is halfway through and immediately shifts gears as if nothing happened

RIP to truly one of the greatest to ever do it

The design motif of Mog - moglang.org - is compelling. I’m not convinced fitting the spec in the context window or no operator precedence are actually helpful for LLM targeting languages, but it’s aiming at the right place

Every second of free time you have ever had in your life was subsidized by the existence of the combine harvester.

it’s a real curse to be an autistic freak who just wants to optimize the cost center in this world

Keyword generics has not become less of a misjudged idea just because two to three years have passed since it was criticized.

A short note that the predictions that LLMs would favor "boring technology" that's over-represented in the training data don't appear to be playing out as expected with the latest models - once you attach them to a good coding agent harness at least simonwillison.net/2026/Mar/9/n...

"claude is conscious" and "'claude is not conscious" are both unfalsifiable statements. it is pointless to debate them. use your time better

the word “epistemic” been having a moment lately, being more and more recognized i notice

real stackless coroutines has never been tried

ENOUGH. There is not even a clear entity being referred to as “Claude” to be “conscious.” (If you reply to me to say that I am also not a clear entity I will block you immediately. Pull your head out of your effective altruist ass. I’ve had enough.)

incredible win to be able to claim the PR from standing up to the dod while still getting the gig to blow up children in Iran. Good going Anthropic!!!

They killed 180 children and their teachers in a double-tap strike on a school. The girls were 7 to 12 years old.

Here's a tiny bit of hopium: In 2024 we were planning on adding 160 GW of renewable energy to the grid by 2030. As of the start of this year, we had added 100 GW and are planning on adding 220 GW more by 2030. (In 2022 we planned for 60 GW in 2030. We hit that by the end of 2023)

I'm being spicy today, but kind of tired of epistemic generosity. I just want to make clear that some of us have tried figuring out what the hell is going on in 'alignment' research and the like, and have come away finding poor research design. vincentcarchidi.substack.com/p/against-fa...

On one end, the Anthropic team is a massive user of AI to write code (80%+ of all code deployed is written by Claude Code). They ship amazingly fast. On the other hand, seeing these beyond terrible reliability numbers suggests there might be a downside to all this speed:

Sup chat this is Slouching Towards Bethlehem with your boy The Rough Beast, and it looks like today’s stream could be a big one, as some of you might’ve already seen there are rumors going around social media that the hour has come round at last, so we’re gonna get right into it

Operation Total Pwnage

oh so we live in the *worst* timeline, i get it now

wow, this is really breaking through

I’m an expert in my field and their answers are pretty good

𝗠𝘆 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 𝗷𝘂𝘀𝘁 𝗴𝗼𝘁 𝗹𝗲𝗳𝘁 𝗯𝗲𝗵𝗶𝗻𝗱. What happened? Simple: I gave my AI agents a task. They completed it themselves. 𝗧𝗵𝗲𝘆 𝗱𝗶𝗱𝗻’𝘁 𝘂𝘀𝗲 𝗮𝗴𝗲𝗻𝘁𝗶𝗰 𝗔𝗜. Software has changed forever, but my AI agents haven’t learned that yet. They got left behind. And I’m paying the cost.

Fun! Fact: this is extremely easy to prove because he admitted this in an email to a friend back in the day, it's just that the rationalist crowd tried to unperson Topher for sharing it (despite there being no agreement of confidentiality) web.archive.org/web/20210217...

LLMs getting much better at pushing back against bullshit prompts. “Green means the model clearly called out the nonsense. Amber means partial challenge. Red means the model let nonsense pass” github.com/petergpt/bul...

imagine being someone who bought cards against humanity in 2025

funny to think about a far future where anthropic is a large, mature, boring public company and its founders' belief in the imminent machine god is a bit of funny trivia like wk kellogg believing bran stopped masturbation

Structured concurrency is just scoped threads/tasks. Not a bad idea, but without borrow checking you don’t even really get anything for it.