What will it actually take to get to AGI? Today we unpack the “jagged frontier” of AI capabilities — systems that can dazzle at PhD-level reasoning one moment but stumble on high school math the next. We look at Demis Hassabis’ timeline and critique of current models, the debate over whether today’s AI really operates at PhD level, and why continual learning and memory remain the missing breakthroughs. We also explore how coding agents, real-world usage data, and persistent context may become critical steps on the road to AGI. Finally, in headlines: lawsuits over AI search, Apple leadership changes, OpenAI’s renegotiated deal with Microsoft, and layoffs at xAI.

Ask this episode

“What did the council decide on item 42?” — answered with the speaker and the moment.