On the one hand, this is amazing. On the other, history suggests that meme-driven revenue tends not to be durable long term. Ethereum went through this, and so did Solana. The test for Robinhood Chain is whether it can turn this burst of speculative activity into something that lasts after the frenzy fades. And now is a great time to prepare for that transition. Attention is on the chain, the community is excited, and users are showing up. Robinhood is also in a uniquely strong position given how central tokenized equities are to the current activity. There is still a lot of white space to build durable use cases around those assets. It will be an interesting saga to follow.

One question I have about the RL space: how durable are businesses built around expert-generated training data? There’s a strange dynamic here. If you recruit experts to train a model in a niche legal domain and the RL works, you may eventually no longer need those experts. In that sense, you’re effectively churning through your own supply side: you spend money finding and onboarding expertise, use it to train the model, then have to move on and acquire a new pool of experts in another domain. That feels different from a traditional marketplace. An Uber driver getting you from point A to point B doesn’t make that driver less useful to Uber. Another rider will still need a ride tomorrow. The act of providing the service doesn’t reduce the future need for the supply side. With RL, the opposite may be true. The better the expert does their job, and the better the model learns from them, the less valuable that same expert may become to the platform over time. There seems to be a few ways to think about this... The bearish view is that these businesses are structurally working through their own TAM. As models improve within a domain, the marginal value of additional expert training falls, while the company has to keep spending to recruit experts in new domains. The bullish view is that this dynamic may be true in the limit, but the limit is a long way away. Models may keep needing new environments, harder tasks, better feedback, and expertise across an expanding set of domains for many years. Curious how people building in the space think about the durability of these businesses. My DMs are open as well.