Wait, vibe-coded software is unreliable?

Oh the sweet irony - all the AI slop can't be copyrighted, so the output of the stealing machine can be taken without it *being* stealing. Probably means that entirely vibe-coded projects have no IP protection. Will be interesting if companies keep humans just for this. apple.news/Av5KfH2DGSmC...

"In 51.6% of cases where someone needed to go to the hospital immediately, the platform said stay home or book a routine medical appointment" SHUT IT DOWN, ffs, how is this even a debate

Jail. Every single one of these fuckers that released a "health" LLM product. Now their racists LLMs not only kill through the US military, but also directly kill their customers. I know in the US, crime is legal, but the @ec.europa.eu should step in!

LLMs don't analyze data. They don't reason. They just predict data based on their training data. So when the US military uses Claude to assess which people to kill, Claude literally makes those decisions based on what it's seen in novels, where they usually murder people with foreign names. 1/4

🎯

We should have never legitimized gambling or outsourced our thinking.

Each one of those three missiles costs 33% more than the annual budget of the national park where I work.

"I think it's worth some empathy to the DoW" This is the guy with the most sensitive data on a billion people. Pro tip: in ChatGPT -> click on your name -> Data controls -> Delete all chats. Something I do every couple of days, so this grifter has less data to hand over to Trump.

The allowlist of Ukraine's Starlinks was created *in a day* and was enforced within weeks. Showing us that Musk always could have helped Ukraine, he chose instead to let them bleed for 4 years for his own political reasons. Musk will rot in hell. apple.news/AzqtGciZzQku...

Based on every crypto bro I've met, it makes total sense to me that Epstein was basically the angel investor of that whole industry for a decade. If Epstein were alive today, he'd be the biggest LLM booster as well. No surprise half of the LLM datacenter co's used to do bitcoin mining 5 years ago.

LLMs are the future ... of comedy

Given how they can only be produced by exploitation of workers from the global majority no system including any of the big LLMs in their production can ever be called "fair". "Fair LLMs" of the size required to do the tasks people want LLMs to do (badly) do not exist.

Requiring constant new money in for existing stakeholders to be paid out, otherwise the whole thing collapses? Why do I feel like I've heard this before en.wikipedia.org/wiki/Ponzi_s...

I wonder how many hundreds of billions lower the valuations of AI companies would be today if more people understood the simple fact that an "AI agent" is nothing more than a loop calling an LLM. An "agent" isn't smarter than a chatbot. It can just do more damage with its stupidity.

We need term limits and psych evaluations for these people, before we hand them keys to trillion-dollar kingdoms. Because right now, we're not sending our best, basically every one of them is an evil sociopath who sooner or later starts kissing Trump's shoes. Can't be a coincidence.

The biggest threat to the LLM bubble is people using their brains. These companies only succeed if you outsource your thinking to them, and they in turn let the highest bidder influence you - LLM ads. That's the only business model that has a chance to pay for their debt obligations. Enjoy.

Anthropic still owes me $3000 for stealing my book

Somebody's lying. Either OpenAI softened on the ban of those two red lines, or the Anthropic fallout was about more than just those two. This is where it'd be awesome to have journalistic institutions *not* owned by the same cabal of billionaires we'd like to be investigated 🫠

All technology is political. It always was. Tech companies are at the center of politics these days, and often have more power than elected leaders. We must scrutinize tech CEOs just as much as politicians. No free pass. No benefit of the doubt. Put them under a microscope and find every skeleton.

This is why I laugh whenever someone says that Dario or Sam are better than Elon. They’re all evil sociopaths. At this point it doesn’t matter if you work for these LLM companies or directly for Palantir. Either way, you picked your side in the history books. apple.news/AxRhzJMHHSKm...

I support tech that shifts power from billionaires to individuals. That's why I love open source. I'm against tech that shifts power from individuals to billionaires. Like LLMs, mass surveillance, and autonomous weapons. Tech is always political. We, the people, decide which tech lives and dies.

So Anthropic is willing to discard its values for a small $200mil Pentagon contract, even though they're valued at $350b and are committing to spending tens of billion on datacenters? Something isn't adding up. What if they genuinely need this cash? 🫢 apple.news/ArPrMpHjGT3y...

People claiming that LLMs will kill SaaS seem to have stopped at the first "S", without knowing what the "aaS" stands for

I made a demo from the outcome of me playing around with @elementary.codes. Now with cute lil’ animations 😍 Cautionary note: I barely know what I’m doing and I’m learning as I go. Browse accordingly 😅 github.com/joforsell/el... #swiftwasm #wasm #swift #swiftlang

One of the big tell-tell signs of a Swift package being AI slop is it being created in 2026, yet using 5.9 as the tools version in Package.swift. Not sure why, but these models love to default to 5.9, instead of at least 6.0. I used to follow new packages on GitHub, these days it's mostly slop.

A random thing I love about the atproto ecosystem: nobody can squat on your username on a new service. Your username is atmosphere-wide. Which also removes that urgency of trying anything new that has a chance of taking off, because, let's be honest, all of us hate not having our usual username 🫢

I also uploaded a template to make it trivial for anyone to deploy a CRUD server in Swift, for obvious reason - it was always supposed to be this easy. railway.com/new/template...

Just a heads up - I'll be switching the underlying infra of my custom swiftlang feeds, so you might see some interruptions in the next 15 mins

I post a lot about tech that I’m skeptical about, so let me cleanse the timeline with some tech I find genuinely awesome. 1. @swift.org (Especially web servers on Linux, and WASM in the browser. Full stack!) 2. @atproto.com (Open social with open APIs? Love it.) 3. @railway.com (Chef’s kiss.)

Am I the only person bothered by the lack of symmetry of the HTTP terminology for "headers" and "trailers"? The opposite of "headers" is "footers". The opposite of "trailers" is "leaders". Help

This is why I haven't gone to the US in 2025, and am not planning to go for the foreseeable future. Also means that if there are any fun conferences, I can't participate if they're in the US. At this point, feels borderline immoral to organize a conference in the US and ask people to risk it all.

I don’t get why Anthropic hasn’t used Claude Code to rewrite their bloated Electron app as native. Should be much easier than a C compiler.

Boosters: "AI slop code taking down AWS is good, actually"

📐Announcing Swift System Metrics 1.0! 🎉 A stable API for process-level monitoring on Linux and macOS. Add it to your service in a few lines, plug into Prometheus or OTel, and start visualizing in Grafana. Contributions welcome! www.swift.org/blog/swift-s...

I guess that means tech CEOs kissed Trump's ass for fun, but definitely not for profit 🫠 It's almost like they would have been better off not being cowards, who knew. apple.news/AkIx53QeiSCa...

Files in repos like AGENTS dot md make results _worse_ and _more expensive_. Glad that me finding these files silly from the start is now supported by research showing they do more harm than good. arxiv.org/pdf/2602.11988

"...the agentic tool... decided to "delete and recreate the environment." An AWS spokesperson [said] that the disruption was brief and attributed it to user error.” --- Now THIS is indeed the future. Employers force AI top-down, and then blame employees for errors. 🫠 apple.news/AJQWqM1ICSrK...

Look ma', I'm running #swift in the browser! Note: totally made up placeholder data 😅 #wasm #swiftwasm #swiftlang

why won’t the left just praise every technology handed down from self-interested billionaires seeking only to enhance their power and profits?! clearly, they’re the ones in the wrong if they can’t get on board 😤

The LLM trifecta, demonstrated by Microslop: 1. Steal someone else's work, no attribution 2. Launder it through an LLM, which makes it much, much worse 3. Ship it to production without a single human reviewing it I can't even be mad though, they gave us the gift of "continvoucly morging" ❤️

With all the hype around the "best coding model", Claude Opus, how much progress has Anthropic made in the past 5 months? Go ahead, guess. In the Agentic Coding benchmark, how much percent did it improve from 4.5 to 4.6? 20%? 50%? Exponential improvement, right? ... Nope. It got worse. 📉

“The playing field is poised to become a lot more competitive, and businesses that don’t deploy cocaine to help them innovate in everything they do will be at a disadvantage,” says Paul Daugherty, Chief Technology and Innovation Officer at Accenture. Hey, it works!

ai;dr AI; didn't read.

Some will put in the work. Others will use that time to instead learn actually useful skills that don't become irrelevant the moment the next minor model update comes out. To each their own.

Slop is the perfect word to describe the low-quality, high-volume, AI-generated content being thrown around. If you don't have time to write it, I don't have time to read it.

I already mentally group products/projects/people this way now. Human creations are original, thought-provoking (especially when I disagree), and expand our horizons. AI slop is, by definition, the most average, thoughtless, unoriginal thing possible. Only the former gets my attention.