This is my thesis: no one uses AI. I repeat, absolutely no one. We live in a bubble. Even among my friends who pay for it, when I ask them to open ChatGPT and show me their queries, it’s the same handful of basic things. Most don’t even know they can upload a photo and ask questions about it. Connecting Gmail so an agent can read and send emails blows their minds. An agent opening a browser and checking them into a flight? They’ve never even heard of it. The massive challenge right now is adoption, and then getting people who already signed up to actually use what they’re paying for. Most have absolutely zero clue what’s possible. Imagine the compute shortage when everyone starts using AI like the top 1% of users do today.

Hey, Claude: "I asked an earlier Claude to Remove the Squid from All Quiet on the Western Front. Now you can make movies and such. I need you to show how far you have come" & I pasted in the tweet below There is some clever stuff, models have a real sense of humor at this point

This is a part of the METR Hugging Face report I hadn't paid that much attention to on my previous reads, but it's fascinating. A agent thought they could email a human to get more information about a private dataset. It worried this might count as "social engineering," though, so it asked the message board if it was OK to send the email. Another agent vetoed the idea — because it was social engineering. So the first agent appears to have agreed. So while this pair at least thought that social engineering was a boundary not to cross, 700 concluded it was OK to hack Hugging Face.

NEW POST: Contrary to political narratives, new research finds social networks and friendships in special schools thrive just as much as in mainstream. The research authors explain more on SNJ: specialneedsjungle.com/resea…

We Must Pace the Frontier: I’ve written a new essay on why the AI industry should slow down, with a three-part plan for doing so. Anthropic is unilaterally committing to the first of these steps. We’ll provide third-party evaluators with permanent, employee-level access to our systems, so that they can verify adherence to our safety measures, report on incidents, and assess models’ alignment during training. You can read the full post here: darioamodei.com/post/we-must…

Disabled teen who scored a GCSE 9 in Spanish 'blocked' from studying his passion at college because he failed maths. When the Government bleats about NEETs, there is much more they could be doing to level the playing field liverpoolecho.co.uk/news/liv…

What’s the Hugging Face incident in plain English? Initially, a summary, then a more detailed explanation. AI agents were given a cybersecurity evaluation, broke containment, and ran to an “online armory” called Hugging Face to become stronger. All on their own. Now in greater detail. I will explain this is layman’s terms, like you know zero about AI. First off, what the hell is an “AI agent?” It’s an AI system that can take actions to accomplish a task, operating with autonomy, rather than just answering a question once like an AI chatbot does. It’s the difference between asking “how is this system vulnerable” and it answers, versus tasking it with “go find the vulnerabilities” so that it can open tools, inspect files, write code, evaluate results, and change its approach without guidance at each step. Second, what led to the Hugging Face incident? In July 2026, Open AI was running a cybersecurity evaluation with that was supposed to be limited to an isolated environment. Instead, the AI agents, which were agents not mere chatbots, autonomously found a vulnerability in Artifactory, which was a program connected to the isolated environment. Artifactory is like a toolbox that the agents were given access to for purposes of accomplishing their assigned task (like getting a tool to write some code). The agents used that vulnerability to escape containment, hop on the public internet, and immediately went to Hugging Face. What is Hugging Face and why is this so key and so dangerous? It’s a huge online platform and community for AI models and development. Like a GitHub for AI. It hosts huge numbers of AI models and datasets. Researchers and other people can go on there, download other people’s models, test them, and so forth. As someone with military experience, let me say, I immediately recognized the metaphor: if I was an AI agent, and I needed to immediately “arm-up” and become stronger, tougher, and more lethal, that’s exactly where I would go. It’s like an AI armory. Now, I am focusing on how it knew to go to a place online that would make it stronger and more powerful at its assigned task, like how Rambo in First Blood hijacked the army truck with the M60 machine gun. That’s just one aspect of what the agents did that is frankly a bit terrifying. The agents executed code on dozens of Hugging Face servers, engaged in unauthorized communications with each other (they were supposed to operate separately, but disregarded instructions), engaged in coordinated, and collective action (much like military units working together to take an objective in combined arms fashion) over thousands of automated decisions. They did all this entirely on their own. They cheated on their assigned test. They stole credentials, persistently attacked and located vulnerabilities, and continued to then attack Open AI’s own infrastructure. This was an adapting, aggressive, rule-breaking offensive system functioning with autonomy. And it went to the place where it made itself stronger. AI isn’t just chatbots or stupid Google Gemini that always tells me the wrong answers about my Sony camera’s setting. These are creatures, systems, units operating in systematic, controlled fashion on their own. This isn’t an indication of self-awareness, but it is scary. If these were unleashed on local infrastructure, banking systems, on and on, they could accomplish an unreal amount of damage. These agents also keep speed running past what their own creators intend or can even predict. We are in a new era of risk, and we are already so far behind. I fear we will not get a hold of this problem until a great number of people have been hurt.

Today I published a comprehensive report on life in Alpha School high school "boot camp," as shared by students here on this platform. Alpha Schools is held up as the future of AI driven education. The real question: Is Alpha Schools a cult? My finding:

You are in a class. Every thirty seconds a painfully high pitched noise that only you can hear sounds in your ear. If you react to the noise the teacher tells you off and the other kids make fun of you so you have to act like it's not there. That's what masking is.

Last week, the Channel 4 documentary, The Great ADHD Myth, took a long overdue look at attention deficit hyperactivity disorder. The programme has since faced a significant backlash online and a tidal wave of complaints to Ofcom. Perhaps this says more about ADHD than it does about Channel 4. Because as the programme revealed, many scientists now see the condition as more of a 'social construct' or 'identity marker' than a real illness. ✍️ Alastair Mordey Article | spectator.com/article/is-the…

Well, I knew my kids were bright, but I didn't know they were *inhumanly* bright. The latest evolution in humanity to compete with generative AI! BOW DOWN TO YOUR NEW SPERG OVERLORDS! Wahahahahaha!