Sebastien Bubeck of OpenAI says “yes, nonsofic groups exist”—as an example of “many new beautiful results” from Astra, next major OpenAI model. OpenAI published a page and PDF detailing ten advances across pure math and theoretical computer science. The arguments were generated by an internal version of Astra (total compute cost ~$2,000 at their Sol …
When you use an AI to write code, you pay for tokens — chunks of text the model reads and writes. Frontier models charge a lot per token. Cheaper models charge a fraction. Think of building a house. An architect charges $500 an hour; a bricklayer charges $40. Both are essential. But you’d go broke …
Gavin Baker says anything that increases competition and compresses margins at the model layer (foundation models) is net positive for every other layer in the AI stack — power, semiconductors, hyperscalers/neoclouds, and even software. A world with only 2–3 dominant high-margin frontier labs is bad for the rest of the ecosystem because those labs become …
High reasoning seems to be the sweet spot of value for Anthropic Fable and OpenAI GPT 5.6. Key GTP 5.6 Sol Problems Theo Highlights That Are Similar for Anthropic Fable GPT-5.6 Sol is excellent (strong agentic coding, long-horizon tasks) but much more persistent—it keeps going instead of stopping early like GPT-5.5, leading to unpredictable and …
Theo spent six figures in tokens testing OpenAI’s 5.6 Sol model to see whether if its better than Fable and what OpenAI have done to make it even better than 5.5. They breakdown why they both moved their agents to Linux boxes, how to actually burn $65k on a single loop, and the Codex vs …
SpaceX recently completed the largest IPO ever, raising more than $85 billion and debuting at a ~$1.77 trillion valuation. Shares have since pulled back (a common post-IPO pattern, especially in volatile markets), but the raise was enormous. xAI is part of that story because of the earlier integration. FEb 2026, SpaceX acquired xAI and the …
Theo is an AI practitioner. He and his team use the AI to code for their own startup. He personally uses over $500/month of API compute and members of his team use $2000/month. He indicates the shift for Anthropic was Claude 4.5 becoming really useful. The pricing then shifted and they bumped everyone up by …
XAI gets to about breakeven by renting half of their AI Data center and Anthropic can serve its models and more money and can get to a better IPO. Both get to IPO before OpenAI. Both stronger against OpenAI. OpenAI IPO is hurt. Anthropic is already at pre-IPO valuation of $1.2 trillion. SpaceX XAI will …
OpenAI GPT-5.5 tops the Artificial Analysis Intelligence Index by 3 points, breaking a three-way tie with Anthropic and Google. There is indication that is faster and has some good results but some results are mid. More time is needed. It is a good model but it does not appear to be groundbreaking. OpenAI topping five …
xAI’s Colossus 2 (MACROHARD) now has 7 models in training at once. Grok Imagine V2 2 variants of 1-trillion-parameter models 2 variants of 1.5-trillion-parameter models 1 variant of a 6-trillion-parameter model 1 variant of a 10-trillion-parameter model SpaceXAI Colossus 2 now has 7 models in training: – Imagine V2– 2 variants of 1T– 2 variants …
Anthropic signed a new agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity that we expect to come online starting in 2027. This significant expansion of their compute infrastructure will power our frontier Claude models and help us serve extraordinary demand from customers worldwide. The run-rate revenue has now surpassed $30 billion—up …