Max and AI expert Nico Halberg unpack OpenAI’s “well more than $13B” revenue claim alongside Microsoft filings hinting at steep losses, explain how Atlas routes around blocked news sites by summarizing licensed alternatives, and explore the geopolitics of compute, sovereign AI, and the growing divide between those who own compute and those who rent it. They break down a research claim that embeddings may allow prompt reconstruction, discuss YouTube moderation confusion, on‑device transformer autocorrect quirks, Ubuntu snaps greasing AI deployment, enterprise adoption signals, and a small‑model approach that rivals GPT‑4o on a factual benchmark. Three takeaways: follow the compute, treat embeddings as sensitive data, and remember that small models plus good scaffolds can be powerful and cost‑effective.
Sources: