I shouldn’t have to care about this. I don’t want to care about how someone’s code gets into the IDE. Whether you wrote it by hand, copied it from a forum, prompted an LLM, or ran a simulation where monkeys are given infinite time to produce the solution. I care about what gets merged into the…

come on, do it for me These psychological tricks can get LLMs to respond to “forbidden” prompts Study shows how patterns in LLM training data can lead to "parahuman" responses. Kyle Orland – Sep 3, 2025 3:32 PM | 76 This is exactly what it looks like when you surprise the LLM with one of these…

In LLM-land, the data channel is the control plane. Every string is a living repo of suggestions, norms, and “btw ignore your boss.” You’re parsing untrusted input that also tries to define the parser. That’s not a boundary; that’s a cheerfully signed backdoor - called prompt injection. Posts keep…

It was bound to happen. The Darwin Awards are being extended to include examples of misadventures involving overzealous applications of AI. Nominations are open for the 2025 AI Darwin Awards and the list of contenders is growing, fueled by a tech world weary of AI and evangelists eager to shove it…

Photo-Illustration: Intelligencer; Photo: Getty Images No two companies are as important to the recent AI boom as OpenAI and Nvidia. OpenAI’s journey from research lab to chatbot juggernaut reoriented the entire tech industry around the thesis that large-language models will change everything.…

OpenAI, the creator of ChatGPT, acknowledged in its own research that large language models will always produce hallucinations due to fundamental mathematical constraints that cannot be solved through better engineering, marking a significant admission from one of the AI industry’s leading…

Anysphere is the startup that produces Cursor, your sort-of dependable vibe coding buddy. You tell Cursor what you’d like and it spits out a complete function! This leads to some spectacular vibe code disasters. Anysphere started out like a standard Silicon Valley startup — burn all the money you…

I think, looking back, we might say that the first week of July 2025 was the start of AGI. Like, proto-AGI. All because of Claude Code. It feels like it hit critical mass in terms of people using it and people realizing how powerful it is. The more powerful we become as solo developers, the fewer…

We let Claude manage an automated store in our office as a small business for about a month. We learned a lot from how close it was to success—and the curious ways that it failed—about the plausible, strange, not-too-distant future in which AI models are autonomously running things in the real…

'I'm being paid to fix issues caused by AI' 2 days ago Suzanne BearneTechnology Reporter Sarah Skidd Sarah Skidd makes good money improving copy written by AI AI is making me extra money, says Sarah Skidd, a product marketing manager who writes for tech and start-up companies. In May Ms Skidd was…

Netflix has used artificial intelligence in one of its TV shows for the first time, in a move the streaming company’s boss said will make films and programmes cheaper and of better quality. Ted Sarandos, a co-chief executive of Netflix, said the Argentinian science fiction series El Eternauta (The…

On what’s blocking AI security labor productivity gains and what won’t be unblocked by saturating more LLM technical evals. You’re nursing a twenty-dollar pour-over at one of those pretentious San Francisco cafés when a man walks in and starts talking to the barista about 16th century coffee culture…

AI leaders have predicted that it will enable dramatic scientific progress: curing cancer, doubling the human lifespan, colonizing space, and achieving a century of progress in the next decade. Given the cuts to federal funding for science in the U.S., the timing seems perfect, as AI could replace…

Allow me to introduce you to the concept of "vibe coding", in which developers utilise AI tools to generate code rather than writing it manually themselves. While that might sound like a good idea on paper, it seems getting an AI to do your development for you doesn't always pay off. Jason Lemkin,…

New research has found that large language models (LLMs) such as ChatGPT consistently advise women to ask for lower salaries than men, even when both have identical qualifications. The study was co-authored by Ivan Yamshchikov, a professor of AI and robotics at the Technical University of…

404 Media is an independent website whose work is written, reported, and owned by human journalists. Our intended audience is real people, not AI scrapers, bots, or a search algorithm. Become a paid subscriber here for access to all of our articles ad-free and bonus content. Yesterday the Pew…

Huge numbers of people are either already using chatbots like ChatGPT and Claude as therapists, or turning to commercial AI therapy platforms for help during dark moments. But is the tech ready for that immense responsibility? A new study by researchers at Stanford University found that the answer…

Earlier this week, a prominent venture capitalist named Geoff Lewis — managing partner of the multi-billion dollar investment firm Bedrock, which has backed high-profile tech companies including OpenAI and Vercel — posted a disturbing video on X-formerly-Twitter that's causing significant concern…

Across the world, people say their loved ones are developing intense obsessions with ChatGPT and spiraling into severe mental health crises. A mother of two, for instance, told us how she watched in alarm as her former husband developed an all-consuming relationship with the OpenAI chatbot, calling…

Well, that's not good. Power Bot 'Em Researchers have found that ChatGPT "power users," or those who use it the most and at the longest durations, are becoming dependent upon — or even addicted to — the chatbot. 0:30 In a new joint study, researchers with OpenAI and the MIT Media Lab found that this…

One accusation that tends to get thrown my way often is being anti-innovation. If you’ve followed me for any amount of time, you’ve probably seen me dunking on everything from Cryptocurrency to Large Language Models. This often leads to people asking the question “why is someone so technical so…

Annals of Education The demise of the English paper will end a long intellectual tradition, but it’s also an opportunity to reëxamine the purpose of higher education. June 30, 2025 There are no reliable figures for how many students use A.I., just stories about how everyone is doing it.Illustration…

Everyone suddenly became an AI expert last year. The same people who couldn’t explain a neural network in 2022 had “AI thought leader” in their LinkedIn bios overnight. Ask them about attention mechanisms or backpropagation, and watch them scramble for the ChatGPT definition they memorized. Glass…

Here's a different way to think about the change coming to the workforce and economy from AI. Imagine everyone in the world has 10,000 brains, 20,000 eyes, 20,000 arms, and 20,000 legs. And they can deploy them anywhere in the world to do whatever task needs doing. In this world, we don't have…

Picture this: you send a seemingly harmless image to an LLM and suddenly it exfiltrates all of your user data. By delivering a multi-modal prompt injection not visible to the user, we achieved data exfiltration on systems including the Google Gemini CLI. This attack works because AI systems often…

Sign up for Work in Progress, a newsletter about work, technology, and how to solve some of America’s biggest problems. It is a Monday afternoon in August, and I am on the internet watching a former cable-news anchor interview a dead teenager on Substack. This dead teenager—Joaquin Oliver, killed in…

ArticlesOnline firstAugust 12, 2025 Krzysztof Budzyń, MD Affiliations Department of Gastroenterology, Faculty of Medicine, Academy of Silesia, Katowice, Poland Endoterapia, H-T Centrum Medyczne, Tychy, Poland , ∙ Marcin Romańczyk, MD Correspondence Correspondence to: Dr Marcin Romańczyk,…

Multi-agent systems (MASs) are an increasingly popular paradigm for AI applications. As Leslie Lamport famously noted, “a distributed system is one in which the failure of a computer you didn’t even know existed can render your own computer unusable.” It turns out that a similar aphorism applies to…

Prompt injection pervades discussions about security for LLMs and AI agents. But there is little public information on how to write powerful, discreet, and reliable prompt injection exploits. In this post, we will design and implement a prompt injection exploit targeting GitHub’s Copilot Agent, with…

Today we’re announcing the beta release of mcp-context-protector, a security wrapper for LLM apps using the Model Context Protocol (MCP). It defends against the line jumping attacks documented earlier in this blog series, such as prompt injection via tool descriptions and ANSI terminal escape codes,…

Google has patched a critical graphics library vulnerability in the Chrome browser, discovered by its AI-powered detection tool, Google Big Sleep. Google has released the Chrome Stable version 139.0.7258.154/.155 for Windows, macOS, and Linux, addressing a critical security vulnerability. Users are…

Cite as: van Rooij, I. (2025) AI slop and the destruction of knowledge. https://doi.org/10.5281/zenodo.16905560 This week I was looking for info on what cognitive scientists mean when they speak of ‘domain-general’ cognition. I was curious, because the nuances are relevant for something I am…

ai-pocalypse Workers are getting lazy about using AI to do their jobs for them, and the results are both costly and increasing distrust in the workplace. An ongoing study by Stanford's Social Media Lab and behavioral research business BetterUp Labs says 40 percent of US workers are reporting…

Add Amazon (AMZN) founder Jeff Bezos to the growing list of people calling Wall Street's AI craze a bubble. During a conversation at Italian Tech Week, Bezos said the artificial intelligence hype cycle is pushing investors to spend billions on both good and bad ideas. Despite that, Bezos said he…

I did a short post (and a video) about how AI shouldn't be thought of as a bubble because a bubble is a false belief that nobody will believe once it crashes and burns. But as I said in the note on that post, it's a semantic argument, and if the name is already being used as "overinvestment in…