I’ve heard several OAI people say there are unfixable bugs in hardware that are exploitable by AI models directly speaking machine code so even if we could solve software that’s not enough. We need defense to move faster than offense if that’s even possible we may just have a new normal that requires adapting to insecure software (bank transfers require going into a branch and taking a dna test 🤣😬)

We placed the fruit fly brain inside Odyssey-3, a frontier world model. The lil’ guy is now flying around totally imagined, simulated worlds all by itself 🫶

Peter Thiel explains why aliens with faster-than-light travel must be demons or angels: only a demonic totalitarian mind meld or perfect angelic altruism could keep any one of them from launching a warp drive weapon "...If you have faster than light travel, if you have a warp drive, which is probably what you really need to cover interstellar distances, you know, what that means for military technology is that you can send weapons at warp speed and they will hit you before you see them coming. And there is no defense against a warp speed weapon." "And so it tells us that I think that if you have faster than light travel, there's something really crazy that has to be true on a cultural, political, social level. And there may be other solutions, but I'll give you my two." "You have a demonic totalitarian control of your society, where it's like you have like parapsychological mind meld with everybody, and no one can act independently of anybody else. No one can ever launch a warp drive weapon." "And then I think the other version, socially and culturally, is they have to be like perfectly altruistic, non-self-interest. They have to be angels." "It's not that they might be demons or angels, they must be demons or angels if you have faster than light travel. And both of those seem pretty crazy to me."

Software is dead. the new frontier is Data

Introducing Kardashev-0.7, the world’s first trained swarm made of 32 distinct models Trained with RL for Population Scaling (RLPS), 32 models organically develop specialization & complementary capabilities, delivering frontier performance at: - 0.007x ~ 0.02x of the inference cost - 0.03x of the required memory Civilization advances through different minds specializing and working together. We’re bringing that principle into AI @BanburyRoadAI, we’re scaling intelligence by model count, toward civilizations of models that learn to build on one another

SemiAnalysis numbers imply OpenAI is selling Cerebras-powered Ultrafast inference at ~$200M per megawatt per year. People are re-running the maths hoping for a typo. We're not surprised. A megawatt is a megawatt. What it earns depends on the silicon inside it. Here's how we think about it: Near term, revenue per megawatt wins. Power is the scarcest input in AI, and whoever monetizes each MW best gets the next MW. Long term, gross margin per megawatt wins. Prices compress, competitors arrive, and only the operators with real cost and pricing advantages keep the spread. Cerebras is built for both. Speed is the product. A frontier model at 750 tokens per second is something GPU clouds haven't publicly matched, and customers like Jane Street pay a premium for it. Premium pricing on a contracted, scarce resource is how you protect margin when everyone else is racing to the bottom. That's why we're extremely bullish on Cerebras, and why we're buying as many megawatts as we can. Premium inference unlocks the most delightful user experience in AI. No more waiting, waiting, waiting, staring at little thinking animations while the model works in the background. Answers that keep up with your thoughts. Our customers are even more excited than we are. That's General Compute.

Ratio of real pricing per 1M tokens with a 95% cache hit rate to the fair price estimated from model benchmark performance. Lower is better. Top 3 models are DeepSeek V4.1 Flash, GPT 6 Luna and Sonnet 5.5. They are appear to be heavily underpriced relative to their benchmark performance.