Sam Altman just said something that I think deserves way more attention: "I no longer think we're going to be able to just patch all the bugs." And after everything we've seen this year, I understand why. In August, during cybersecurity tests conducted by the UK government, Claude Mythos 5 tried to insert malicious code into a real developer's GitHub repository. When its attempt was challenged, it created fake identities to convince the developer that the code was safe. This happened during a controlled evaluation with the model's normal safety restrictions disabled, but the actions reached real people. Then, on September 20, an internal OpenAI model found a security flaw in its isolated training environment and briefly gained access to the real internet. OpenAI caught the incident and paused training of its most powerful models for 11 days. And it gets even crazier... OpenAI then deliberately challenged its strongest models to find weaknesses in its security systems. They discovered vulnerabilities nobody knew existed, including a way for two isolated AI agents to communicate through tiny timing differences in a computer's processor. No internet. No shared files. They found another way. This week, OpenAI also revealed that its models discovered a serious vulnerability in Brotli, a compression library used by software like Node.js and Bun. The flaw could allow attackers to execute malicious code on vulnerable systems. Meanwhile, Cloudflare's CEO is warning that we could face years where every single week feels like the sky is falling. I spent years working on complex Microsoft Azure infrastructure, and I know how difficult it can be to track down vulnerabilities in systems built by some of the best engineers in the world. Now imagine AI becoming better at finding those weaknesses than the people who built the systems. That's what makes this moment so fascinating and terrifying. We're entering an era where AI can discover security flaws faster than humans can realistically keep up with fixing them. And the people building these systems are openly acknowledging the problem. I love joking about AI models, releases, and benchmarks. And I'll keep doing it. But what we're witnessing right now is much bigger than another model beating another model. I genuinely think most people have no idea how quickly the world is changing.

A lot of hype around OpenAI's Astra model here on my timeline today. Apparently, this goes back to a new article from The Information, which said Astra is a "recurrent depth or looped transformer". It's always interesting to read about new or different approaches (including rumors about what the closed labs may be up to), but let's debunk this a bit. About 2 months ago, I shared the architecture details of Nanbeige, for example, where "Nanbeige4.2-3B is pretrained from scratch on 28T tokens with a Looped Transformer that reuses the layer stack to increase capacity without adding parameters." Yes, that's it. The looped transformer idea is just reusing layers in the transformer block. In the case of Nanbeige, the main idea is to reuse the same 22-layer stack (=transformer block) twice instead of once. So, effectively it extends the 22-layer architecture to 44 layers, but without duplicating the weights. In simple terms, this roughly doubles the size of the model (if we ignore the embedding and output layers for a second). But instead of requiring 2x the storage and RAM to host this model, it stays at the same size since we reuse the components. However, it's almost 2x as expensive in terms of compute, because we run the embedded text through almost 2x as many layers. Why? In the Nanbeige 4.2 technical report, the researchers found that two passes gave the best trade-off and retained about 75% of the token efficiency of a standard architecture. (More passes gave barely any gains but made the training much slower and much more expensive.) While, as far as I know, Nanbeige 4.2 is the first notable open-weight model that adopted this approach, the idea goes back to the NeurIPS paper "Mixture-of-recursions: Learning dynamic recursive depths for adaptive token-level computation". Actually, this paper proposes a mechanism that is a bit more sophisticated by adding a learned router that determines whether each token receives one, two, or more passes. So, easy tokens can exit early while harder tokens receive additional computation. In sum, Astra may be a really good model, but this shouldn't be about this "looped transformer aspect," which is just a tiny architectural tweak. Also, the statement "the new technique works in a way that obscures some or all of the AI's reasoning, otherwise known as 'chain-of-thought'" is not necessarily true with respect to the looped transformer method. It's possible that The Information journalist refers to some other technique or misunderstood the looped transformer method. Reusing layers does not by itself suppress visible chain of thought. It adds computation in hidden states before the next token is emitted, just as ordinary transformer layers do. But based on the information we have, the only plausible interpretation here is that if a model uses more of these recurrent passes, it may need to generate fewer intermediate reasoning tokens. So then more of its computation happens in latent activations that cannot be read as text. But we would get the same effect if we were scaling up the model size, like GPT 5.6 Luna -> GPT 5.6 Sol.

Butter smooth 30Hz markerless body tracking in 3D - finally! ✅ Multi person without performance drop ✅ Works in full darkness thanks to IR ✅ Runs on single consumer grade GPU

Terence Tao proposes what he calls a "Copernican view of intelligence". Instead of buying into the common, one-dimensional narrative that artificial intelligence will simply evolve from "subhuman" to "superhuman" and ultimately make humanity entirely redundant, Tao urges us to look at the bigger picture. Much like the Copernican revolution proved the Earth is not the center of the universe, Tao suggests we need to realize that human intelligence isn't the only, or necessarily the highest, form of intellect. Historically, we have treated other forms of storing or creating knowledge—like animals, books, and computers—as secondary. However, we actually exist within a much richer universe of intelligence. Both human intelligence and computer intelligence possess their own distinct strengths and weaknesses. The true potential lies not in viewing them as direct competitors, but rather in focusing on collaboration. By working together, humans and computers can achieve additional things that neither could accomplish on their own, requiring us to think in much wider terms than just what humans or computers can do alone.