Every year, some government tries to add surveillance to our messaging apps. They just can't resist the power grab

If you're worried about job security because of AI, don't. There will be lots of security patching jobs to do very soon

This is where we need to ban AI. But the anti-ai people never talk about the institutional harms that an algorithm or AI can enact ok behalf of the government

Spiral of Annihilation also happens to be the name of his band

I thought this was obvious to anyone who's used an LLM to edit a document and actually read it afterwards... But now I'm worried people aren't doing the second part

oh good, we're letting the AI choose who gets to eat

This graph should strike absolute terror into every software engineer who is not already preparing for AI-powered security hacks hacks.mozilla.org/2026/05/behi...

The enchanted forests of Finland-based artist Ulla Thynell

always worth a reread

guy that works at the token-selling company recommends using more tokens. news at 11 x.com/i/status/205...

"We should be very, very wary of building a “papers, please” web where everyone’s online activity is tracked and linked to their official identity"

$24B is less than some junk AI startups

So they wired up human brain cells and the first thing they thought of is to put it in a simulation of hell? kuber.studio/blog/Reflect...

by Aman Daharwal #art

please lord let this pretty goth singer screaming metal covers with the power of 1000 banshees be real and not AI

With the RAM shortage, can someone please invent better memory debugging tools? I want to see a flame graph but for RAM usage: hover over any variable and see how much memory it takes, then dive deeper into each of its fields, etc

What's the play here for Suno? Are they trying to gut the live concert business entirely?

This is a painting I made several years ago, Summer Reflections 40x40 acrylic on canvas

feels like AI is 10xing too many -1x developers

this captures every anti-ai take on bluesky

This isn't realistic because it's expecting normies to be creative enough to come up with an idea of what they want to watch. They just want to be fed content. However, a creator making an interactive film/game that can react along a plot guideline, like the holodeck, will be a whole new art form.

Sometimes scientists can be really smart and really stupid at the same time. “For 30 years, we’ve taught students that… receptor choice is essentially random.” Quick check: if you figure something is "essentially random", you actually haven't figured it out yet.

My entire life philosophy summed up in one comment

New breakthrough in World Models allows the AI to actually learn the underlying physics to make vision predictions with a tiny 15M param model, trainable on a single GPU arxiv.org/pdf/2603.19312

so frustrating to read posts like this, which correctly lay out a real problem — the superficial signals of effort we used as a proxy for work quality no longer function — and jump straight to “surely there’s no way out of this one for an entire economy with trillions on the line”

fine dining restaurant with a ball pit playground. just drop your kids off and go have a nice romantic date. like mcdonald's but with edible food

expect many such breakthroughs, scientific and mathematical

"AI doesn't work"

How is this on by default? If you can't handle rudeness, then maybe the internet is not the right place for you

Ways to resist AI - Meditate in the forest barefoot - Invent a new genre - Tend a garden - Join or start a mutual aid group - Send a handwritten letter - Tell a bedtime story to a child; ask them what happens next - Hug someone What's yours?

"When something goes viral, especially when it’s a woman, people immediately start comparing… Comparison is the easiest way to consume, but the laziest way to understand, and it flattens artists into trends instead of seeing them as worlds." galoremag.com/sevdaliza/

General strike May 1st

Conceptually it is a pretty smart idea to fast-track training a really good computer control model. But it's creepy mainly because Meta is doing it. That's the problem when your company has demonstrated repeatedly it has no morals.

Huge antitrust just landed: Amazon has been price fixing for years. Whenever a price was lower on Walmart, Lowe's, etc, Amazon would email the brand and tell them "fix it". The brand would tell those stores to raise the price again. Those emails just got released. oag.ca.gov/news/press-r...

Alt text always depends on context. Imagine an icon of a computer printer. It could be used on a page of free icons. It could be on a school's computer lab site. Or it could be part of a button that allows you to print the current page. In each instance, the same icon would need different alt text.

Some people on bluesky are wound a little too tight. You reply with a pretty mellow opinion that might slightly counter their take and they act like you just kicked their dog or something

We really are living in the end times

butlerian community planning hearing

Even if you use e2e encrypted messaging like Signal, your iPhone is still saving those messages to an unencrypted notifications database! I wonder if Android has the same vulnerability www.404media.co/fbi-extracts...

Many struggle with writing alt text for charts and other other data visualizations. Amy Cesal's "Writing Alt Text for Data Visualization" hammers home the importance of explaining the chart type, the type of data, and the reason for the chart. medium.com/nightingale/...

Seth McFarlane's Edgar Alan Poe adaptation "The Telltale Fart"

I cannot believe I just spent an hour reading this article about bread. Superb and hilarious writing www.theatlantic.com/magazine/202...

It's like a tamagotchi, but useful!

Based

Many of the internet’s ills are based on widespread corporate surveillance online. This pervasive system, ran by corporations and governments, stifles our speech, and erodes our privacy. But together, we can dismantle it. www.eff.org/deeplinks/2...

love you too, oakland

like, either you are going to pay an AI company five to six figures or you are going to pay a ransomware gang whatever ransomware gangs are charging, and if you take too long to decide it's the second one

Anthropic just admitted to training on the CoT ("The Most Forbidden Technique"). That means the models will be better at hiding their intent from one of our only methods of monitoring for malicious actions. thezvi.substack.com/p/the-most-f...

We're gonna land on that thing! We're gonna build there! And no NIMBY can stop us

Which quadrant are you on?