Conversations with Tyler · Mercatus Center at George Mason University

Conversations with Tyler 2025 Retrospective

December 23, 2025·1 hr·5 clips

Tyler Cowen names the underrated episodes that delivered magical moments with guests like a popular YouTuber.

A retrospective needs an argument. The cold open treats the show's tenth anniversary as both a milestone and a funding problem, with donor levels tied to transcripts, live shows, listener meetups, and direct access. The tone is practical, almost blunt. Tyler names the size of the project, over 250 conversations, then turns to the machinery needed to keep it running. That setup matters later. The main exchange asks why years of circulating arguments have not produced a serious body of work around them. Peer review gets a fair defense. The guest says refereed papers and accumulated literatures can catch mistakes in a way scattered online commentary usually cannot. Tyler presses on the missing piece. If what is needed is a research paper in a recognizable form, he wants to know what could help bring it into being: subsidy, encouragement, or something else. Refine becomes the test case. Tyler asks whether an AI service offering peer style comments could help turn a blog post into something closer to a paper. The guest has not used it, but he is open to the idea. Maybe this product, or a nearby successor, can soon give useful referee-like feedback. The older claim carries more weight. Brian Caplan's distinction between individual papers and literatures becomes the hinge: trust the pileup of judgment more than a single clever document. The trouble is what happens before that pileup exists. The conversation keeps circling the thinness of wisdom when no serious literature has had time or reason to form. The ending widens back out to staff thanks, Mercatus acknowledgments, and a second decade. It is gratitude with the lights on. Less celebration than upkeep: here is what has been built, and here is what still needs structure before it can be trusted.

As heard by us

A decade of sharp interviews, and the live question is whether the field's most durable arguments have ever faced a referee.

Ten years and more than 250 conversations in, Tyler's retrospective becomes something other than a highlights reel: a live dispute over whether AI safety arguments -- some of Eliezer's among them, said to have circulated for fifteen to twenty years -- have ever actually faced…

Read the full review in PlayNext →

Why you'd press play

A ten-year anniversary where the conversation lands, unexpectedly, on whether AI safety ideas have ever been peer-reviewed.

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from Conversations with Tyler