Episode Two

The second episode opens with HBR's research on meeting hangovers: an empty call consumes attention after it ends. Causes include missing agendas, absent decisions, and no influence over the discussion. The hosts suggest advance material, recorded actions, and moving one-way information into asynchronous formats.

The iRobot story concerns a device that becomes incomplete without its vendor's cloud. A customer owns the vacuum but not its application or digital layer. Open protocols and local control offer an escape, although reverse engineering remains an expensive, niche form of insurance.

Dario Amodei's code-generation forecast is examined alongside Anthropic's origins, safety emphasis, and self-supervised learning at scale. The Turing Award for Andrew Barto and Richard Sutton returns to reinforcement learning's forty-year history and the long path from an idea to widespread use.

Crunchbase's report distinguishes unicorn counts from valuations and shows vertical AI products built on foundation models. The final story claims a faster, lower-power Chinese chip without silicon. A laboratory result, the hosts note, does not yet demonstrate manufacturing readiness.