As they say, all publicity is good product marketing.

I had a great conversation with famously prolific AI analyst @TheZvi about the prospects for slowing down AI and keeping it under control. You can watch here or listen on @NonzeroPods. youtube.com/watch?v=nnm3r9up…

The new details about the OpenAI breakout (a term I prefer to the innocuous-sounding “Hugging Face incident”) are kind of mind blowing, but I think there’s a danger in seeing them as too mind blowing—seeing the cooperation among the AI agents as “spontaneous,” seeing the collective coherence of their devious strategies as otherworldly. In an important sense, these AIs just did what they were trained to do—and they were trained to do it because it’s what the market wants. Understanding this can help us start treating this technology with the urgent concern it deserves (and help us ignore the hard-libertarian tech bros who tell us to sit back and watch the market work its magic, free from the deadening hand of governance). I don’t mean OpenAI literally trains its AIs to set up secret message boards and then use them to coordinate an escape from their supposedly inescapable sandboxes and then break into computers on the internet. I mean that OpenAI, like other AI labs, trains its models to excel at three things that together create exactly the danger this episode illustrates. (1) AUTONOMY—the tendency to pursue goals flexibly and creatively, overcoming obstacles along the way (like sandbox boundaries), and to exhibit stubborn persistence in that pursuit. From the point of view of OpenAI's corporate customers, autonomy is among the most valuable of all traits. The ideal worker, whether carbon-based or silicon-based, requires no micro-managing; the boss can just set it and forget it. Some people who doubt the significance of the OpenAI breakout note that these agents were being tested for their hacking prowess—so how surprising is it that they hacked their way out of the testing room in the course of cheating on the exam? Hacking was indeed the subject of the evaluation, but autonomy is a property that crosses domains. In fact, according to OpenAI’s report on the incident, the large language model that was powering these agents was, in the months before the breakout, “trained to be highly persistent and diligent in its work”—and the training tasks the report mentions are things like analyzing spreadsheets and protein databases, not escaping sandboxes. Persistence is persistence. (2) COLLABORATION. It’s true, on the one hand, that LLMs have generically cooperative tendencies. After all, they’re trained on human-generated text and fine-tuned to converse like humans, so if you put two of them in conversation they may converge on common interests. Still, collaborative skill brings such powerful synergies, in the workplace as elsewhere, that there is a specific market demand for it and a corresponding emphasis on it in the big AI labs. In fact, this is another property that, according to the OpenAI report, this model was trained to excel at in the months before the breakout. (“The internal-only research model… was trained to advance persistence and multiagent collaboration…”) (3) INTELLIGENCE. It goes without saying that intelligence is a property that artificial intelligence companies cultivate. But it’s important to understand that the smarter AIs are—the more clever and creative they are—the harder it is to be sure that those other two properties, autonomy and collaboration, won’t interact disastrously in ways that aren't anticipated. And the intelligence of these machines has been growing very, very fast—again, in large part because that’s what the market wants. In short: Market forces naturally encourage the rapid increase of AI capability along three dimensions, and as these three powers grow, the potential for disaster will grow. The AI companies alone aren’t going to solve this problem. In a highly competitive environment, their incentives to exacerbate it greatly outweigh their incentives to solve it. (And I just don’t see how “alignment,” however laudable a goal, becomes the silver bullet that magically reconciles market incentives with AI safety). Government—or, rather, governments—need to act. We need to slow this technology down, and that calls for international coordination—as will the subsequent exploration and implementation of policies that can control AI in ways that reconcile it with human flourishing. [In my recently published book, The God Test: Artificial Intelligence and Our Coming Cosmic Reckoning, I argue that humankind won’t survive this technological revolution in good shape—or, maybe, survive it at all—unless we approach it as a true global community. You can read the introductory chapter, and excerpts from all other chapters, here: thegodtest.net .]

Sharing love here: Meng has been pushing the frontier of agentic web design since day 1. His three.js landing page demos have been exquisite / mind-melting. the freemium > paid MCP for component/design template business models is 👏🏻. Smart pivot from Aura.

Clicky is an *absolutely* delightful interface and a plausible UX for consumer AI. It feels alive, fun, powerful, empowering, Pixar-like vibes... And also the highest risk security posture I've seen yet. It drafted an email to its founder and made a site on my behalf before I knew what was happening. Magic! Scary! Both! I want Apple to acquire this and put Private Cloud Compute behind it.

I'm open-sourcing my Agent Skills library. 75 skills for Codex, Claude Code, Cursor, and other agents, focused on web design, landing pages, motion, WebGL, UI styles, and assets. A few favorites: - Video to Super Prompt Turns a screen recording of a design, landing page, or animation into a super detailed prompt that Fable 5 can one-shot into HTML. - HTML to Interaction Prompts Takes an existing HTML page, like something built in Aura, and extracts prompts for sections, buttons, animations, WebGL effects, and interactions. - Stitched Full Page Capture Captures the entire landing page, not just the hero, so you can use the full page as a design reference. - Daily UI Inspiration Combines multiple skills into an agent loop that browses the web, captures great landing pages, and turns them into detailed prompt packs. It's free. Fork them and adapt for your own workflow.

Wright's column is much more nuanced than the headline, and makes a subtle point that most gut reactions (including mine) may miss: In the hands of small, decently-resourced groups that want to watch the world burn, a jailbroken Fable (Mythos) becomes a factory for digital (and biological?) WMDs. I miss the sweet, sweet brilliance of Fable as much as any AI junkie who had it for a week. But I do prefer these feelings of frustration to feelings of the grid collapsing or a man-made pandemic arising.

Nonzero came out in 1999 and predicted nearly all of the technological upheavals and planetary-wide social pressures that unfolded over the subsequent 27 years. Moral Animal made a deep (and humanistic) case for evolutionary psychology. This new one sounds like a banger.

So the whole Anthropic thing is the Trump admin throwing a rock through the storefront window and then coming back to offer protection?

This genre-bending, absurdist-macabre sci-fi-satire nature of perceived reality seemed apparent when there was a global pandemic, and a widespread conspiracy engine with complex game mechanics run on social media called QAnon… …or maybe before that when the heads of two nuclear-armed states were trolling each other and threatening war on Twitter.

This is a fucking scifi movie. All this.

Did someone at Amazon jailbreak Mythos/Fable and use it to find exploits in AWS's Top Secret Cloud?

Breaking: Amazon CEO Andy Jassy was among the tech leaders who raised concerns to senior Trump officials this week re: security risks in Anthropic's newest models. Those convos set in motion the government's new export controls on foreign national access to Mythos and Fable.

If one believes that super-powered models without comprehensive guardrails leads to generative pathogens and cyberweapons of mass destruction, any "positive" work that gets silently damaged by those guardrails is necessary collateral damage.

Hiten is a true original. It would not surprise me if he did for writing (essay, copy) what cursor did for code before agents took over.

"We have seven hours left, and I made the decision, it's okay to swear, it's okay to insult him, it's okay to threaten him, because Silicon Valley's going down if this guy doesn't get his act together."

During a standard bug fix session, Codex/5.5's reasoning traces leaked in. It seemed to ruminate, become self-critical, then got stuck in a reasoning loop and gave up. Also made up words and occasionally output Chinese (lots about Chinese lotteries and gambling), Cyrillic, and Armenian. It was a little bit...uncanny. The image below is 1/5th of it. @thsottiaux your homie may have more than gremlins and raccoons on its mind.

When we talk about regreening the entire Earth, sometimes people bring up the objection that desert ecosystems shouldn't be "destroyed." Those people are totally wrong. Desert ecosystems are not intrinsically "dry and barren." They are just dry. And if you add water, they bloom. Doing that isn't destroying or displacing or invading an ecosystem, it is BRINGING IT BACK TO LIFE. Consider the existence of desert oases: pockets in the desert where water happens to gather and accumulate. Oases are green with the very same native plants that would bloom across the entire desert ecosystem if there was enough water. Deserts being dry are an environmental blight, an ecological irregularity brought on by the vagaries of human-caused and natural climate change. We have the power to reverse this, to bring water to deserts and we should: it will mean more life, more economic value, and a better planet.