Andreessen Horowitz's house podcast. Partners and outside guests on AI, crypto, defense tech and how companies actually get built, with the firm's own portfolio never far from the frame.
There is no arc to follow. The feed publishes close to daily and mixes formats freely: conference-stage recordings, cross-posts of partners appearing on other shows, portfolio-company interviews, and standalone research conversations. No episode depends on an earlier one, so chronological order buys you nothing and costs you relevance.
In order. If you only listen to one, make it the first.
Benedict Evans concedes that coding is AI's first real product-market fit, citing GitHub Copilot's run rate going from about $9 billion in 2023 to $47 billion by early 2025. Then he argues that win has become tunnel vision, because nobody has answered who captures the value or what the rest of the economy does with any of it.
Barbara Liskov traces her route from the CLU language to Argus, whose "guardian" objects used two-phase commit so that a computation spanning several machines either finished completely or left no trace. It is the clearest thing in the archive on where replicated systems came from.
Vishal Misra makes the case that in-context learning is Bayesian updating, backed by "wind tunnel" experiments where transformer outputs matched the exact posterior. His answer to what is still missing: these models do correlation well and causation badly.
Markus Villig started Bolt at 19, signing up Estonian taxi drivers street by street, and has raised roughly $2 billion against the $24 billion Uber raised before its IPO. He is unusually direct about the failed detour into selling fleet software to the taxi companies themselves.
Altman's test for when to quit a product is "most of the time when stuff works, it works fast." Lattice pivoted three times before performance reviews stuck, and that version was sold off designs while his co-founder was still writing the end of the review cycle.
Which is the real problem with "where do I start" lists. Each episode above links to its summary on this site, free and without an account, so you can find out which of the five is worth your evening before committing to any of them.
The Marc Andreessen and Ben Horowitz macro two-handers are the most heavily promoted episodes and the fastest to expire, because each one is pegged to whatever was in the news that month. The portfolio-company launch interviews age the same way.
The back-catalog problem is the same problem as the queue problem. Listening to summaries instead of full episodes and fitting long shows into a commute cover two other pieces of it.