Joseph Rosztoczy

It’s time to demystify Mythos. Mythos is not magic. It’s not a doomsday device. It’s the first of many models that can automate cyber tasks (just like coding). OpenAI’s GPT-5.5-cyber can now do the same. And all the frontier models (including those from China) will be there within approximately 6 months. It’s important to recognize that these models do not create vulnerabilities; they discover them. The bugs are already in the code. Using AI to discover and patch them will actually harden these systems. The leap from pre-AI cyber to post-AI cyber means that there will be a big upgrade cycle. After that, however, the market is likely to reach a new equilibrium between AI-powered cyber-offense and AI-powered cyber-defense. Obviously it’s important that cyber defenders get access before cyber attackers. That process is already underway but needs to happen quickly (see point above about Chinese models). Unlike Mythos, GPT-5.5-cyber appears not to be token constrained so it may be the first cyber model that defenders actually get to use.

Suddenly standing alone in the room, I begin by imagining humanity banding together and blue winning in a landslide, and I feel a rush of pride. Red is the genocide button. Blue is the “save humanity from this nightmare” button. I know what kind of person I am. As my hand hovers over the blue button, I can’t help but imagine a gun pointing at my head with a bullet in one of the chambers. I feel a surge of fear shoot through my body. Then I think about all the other people staring at the blue button and thinking the same thing. Surely some of those who initially decided to press blue will succumb to the fear. It starts to feel like a gun with two loaded chambers. A stronger pulse of terror. The more I think about it, the more I worry about other people thinking about it. My heart races. Then I look at the red button—a gun with no bullets in it. A glorious feeling of relief washes over me. Will I hate myself forever if blue wins because enough others were better and braver than me? But don’t I owe it to my family to protect myself? One vote won’t change anything anyway, right? It’s all irrelevant because the mammal I live in has already made up its mind. I wince and press red.

Everyone in the world has to take a private vote by pressing a red or blue button. If more than 50% of people press the blue button, everyone survives. If less than 50% of people press the blue button, only people who pressed the red button survive. Which button would you press?

Using Codex is 50% coding 50% pressing the Update button The Codex team really never sleeps

I asked Jensen: “2 out of the top 3 models in the world, Claude and Gemini, were trained on TPU. What does that mean for Nvidia going forward?” After a long technical back and forth about what the right accelerator for AI looks like (see full episode), Jensen lays down the gauntlet: Jensen: "Nvidia's computing stack is the best performance per TCO in the world, bar none. Dylan's InferenceMax is sitting out there for everybody to use, and TPU won't come, Trainium won't come. Nobody wants to show up." “I'm not offended by other people using something else and trying things. If they don't try these other things, how would they know how good ours is?” “There are always big claims, but look at the number of ASICs that have been canceled. Just because you're going to build an ASIC, you still have to build something better than Nvidia. It's not that easy building something better than Nvidia. Nvidia's got to be missing something seriously, because of our scale and our velocity.”

The three states of modern being: AI Euphoria, AI Psychosis, and AI Psychosis Psychosis.

Starting tomorrow at 12pm PT, Claude subscriptions will no longer cover usage on third-party tools like OpenClaw. You can still use these tools with your Claude login via extra usage bundles (now available at a discount), or with a Claude API key.

I feel like the wisest piece of advice for anyone to adopt socially is just to apply this mindset to everything