xAI released Grok 4.5 and OpenAI shipped GPT-5.6 in the same week, and this episode breaks down where each model actually lands on coding, reasoning, and benchmarks. The more consequential news may be OpenAI's full-duplex voice mode, which allows simultaneous speaking and listening rather than turn-based interaction. That shift in interface design signals where human-AI interaction is heading, and the hosts treat it as more than a feature update.
China's AI labs are actively building their own chips to cut dependency on NVIDIA and restricted imports. That story sits inside a broader hardware decoupling effort with long-term supply chain consequences the episode examines with specific context. Cursor's deal economics also get scrutinized, with the hosts making the case that cheap intelligence at scale changes how developer tools get priced and sold.
The back half covers Meta's new image and video models, ByteDance's SeeDream image generation model, and Cloudflare's payments API designed for micropayments and AI agent access. The Cloudflare piece is worth your attention because it points toward an emerging infrastructure layer for agent-to-agent commerce. The full episode is 28 minutes and the timestamps are tight enough to skip directly to what you care about.
[WATCH ON YOUTUBE →]