Ed Zitron's show, arguing that the economics of the AI industry do not work and that the current spending will end badly. It is a case for the prosecution rather than a survey, and it repays listening to it as one.
The show is built out of numbered series rather than a continuous run: a four-part dot-com comparison, a four-part Enshittifinancial Crisis, a three-part NVIDIA sequence. Each series is self-contained and restates the argument with newer figures, so the right move is to pick a series opener rather than the beginning of the feed.
In order. If you only listen to one, make it the first.
The best-sourced episode in the feed. Economist Paul Kedrosky argues that AI is close to invisible in productivity data but dominates non-residential fixed investment, and that data-center construction was the largest single contributor to US GDP growth in three of the four quarters of 2025.
Part one of four, and the clearest statement of the show's central claim: that the dot-com comparison flatters AI. Fiber laid in 1999 stayed useful for decades, whereas GPUs depreciate on a schedule that the multi-decade data-center leases were not written around.
Worth hearing because Zitron spends a good part of it dismantling a viral article that alleged $610 billion of circular-financing fraud at NVIDIA, and concludes the claims were fabricated. It marks where he thinks the critique of the company overreaches.
Matt Rosoff joined CNET in 1995 as employee 42 and gives a first-hand account of the last bubble, from funded startups with no working product to the circular fiber-optic deals at WorldCom and Sienna. The AI parallels get drawn, but the history holds up on its own.
Zitron extends Cory Doctorow's enshittification model with a fourth stage, in which a platform starts sending misleading signals to its own investors. Meta is the worked example, covering its scam-ad revenue exposure, off-balance-sheet data centers and video metrics.
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 Monologue episodes run 5 to 20 minutes and react to that week's news, so they date quickly. The CES 2026 coverage runs to ten parts and roughly 20 hours of convention-floor material, which is not a first listen.
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.