Many longtime fans of my Hexapodia podcast with Brad DeLong have been begging us to bring it back. Well, here it is! Hexapodia will now be a series of Substack Live conversations that you can tune in to, and which will be recorded as podcasts and sent out as emails.
Our first episode was a little rocky, as I got the time wrong and ended up arriving 20 minutes late. But Dan and Brad are both brilliant people, so you can watch them talk for a while before I show up! The topic of our discussion was the Anthropic IPO, inspired by my post about it last week:
I give the bull case for Anthropic, while Dan and Brad are both more skeptical. I think this quote sums up my side of the argument:
My bull case [for Anthropic] is not envisioning a world where Anthropic rules all businesses, rules 100% of every industry, or close to it. I’m also not envisioning a case where AI replaces human beings in their jobs.
I’m envisioning a change in the capital structure of the efficiently producing economy. I’m envisioning…a world where in order to succeed in any industry, you must create a tie-up with a frontier AI company that offers you access to an internal proprietary frontier model that will be months ahead of any model distilled from it. And…that doing this will require not purchasing services…at arm’s length from that company, but tying up your capital structure with that company so that you become in some sense a subsidiary of that company. I am envisioning that.
So I’m saying that instead of something where…AI is like oil and everyone has to buy oil or the things made downstream from oil…instead of that sort of arm of length world, I’m envisioning a world in which every company in order to succeed must become a subsidiary of Anthropic, OpenAI, [or] Google…and Anthropic is simply the biggest of those fish in a oligopolistic ecosystem of a few conglomerates competing against each other via their little subsidiaries in every single industry at the same time.
Interestingly, some economists at Google Deepmind just released a paper with a model in which this sort of thing happens. From the abstract of Imas, Koh, and Legg (2026):
When will advances in AI lead to economic transformation? We develop a model of how intelligence…matters for organizations….AI expertise is reproducible while humans’ are in fixed supply, and AIs may not face the same human limits on coordination….When AI is more capable than the average human, organizations of complementary humans can nonetheless outperform AI. Increases in AI competence can change the composition of workers within firms and decrease firm size. As AI coordination improves, however, this can break the Coasean ceiling on firm size and the market structure undergoes a phase transition toward large, concentrated AI-heavy firms. [emphasis mine]
I was envisioning a slightly more complex version, in which humans and AI are complements, where humans have private data that AI is better at using, and where one big conglomerate enters every industry but retains human workers. But I think Imas et al.’s model gets at some of the basic dynamics here.
Financial disclosure: I have no current position in Anthropic, but unless something major changes, I will probably buy Anthropic stock when it IPOs. I own stock in TypeSafe, an Anthropic competitor.
