OpenAI released ChatGPT 5.5, and the hosts argue it outperforms Anthropic's Claude Mythos on coding and problem-solving benchmarks. The episode runs through inference costs, pricing structure, and real demonstrations including a space mission simulator and a 3D dungeon game built with the model. The math capabilities segment at 13:01 is where the discussion gets technically specific.
The episode also covers SpaceX's partnership with Cursor at the 15:30 mark, a detail worth attention given what it signals about AI tooling adoption at serious engineering organizations. Amazon's continued investment in Anthropic and the rise of Chinese open-source competitors get their own segments, framing the competitive landscape beyond the usual US-centric view.
The reason to watch the full episode is not the ChatGPT 5.5 announcement itself, which is already covered elsewhere. It is the side-by-side capability comparison with Mythos and the discussion of what open-source pressure means for frontier model pricing. The presidential comments on AI at 30:44 close out the episode with a policy angle most tech coverage ignores.
[WATCH ON YOUTUBE →]