I've spent the last four years inside protocols making their most important decisions.

Not as a founder. Not as an investor. As the person in the room who figures out what the company actually is, then tells that story to the world.

At LayerEdge, that meant driving the company through its hardest pivot — from Bitcoin L2 to ZK aggregation infrastructure — writing every external word, closing every partnership, and rebuilding the narrative from scratch when the product changed underneath us.

Now I'm in Dubai, thinking about something I don't think enough people are taking seriously yet:

What does commerce look like when the buyer is an AI?

Not crypto payments for humans. The infrastructure for autonomous economic actors — the identity, the rails, the trust frameworks — for a world where AI agents spend money.

Working inside ZK infrastructure, you develop a specific paranoia about edge assumptions — the things a system takes for granted about whoever's on the other end of it. For two years, every system I worked on assumed a human at the edge. A person with a wallet, a jurisdiction, a liability to bear. ZK proofs, account abstraction, rollup finality — all of it designed for the case where a person decides to sign.

Then language models started acting as agents, and the assumption broke quietly. Not with a warning. Just: one day the entity initiating the transaction might not be a person, and everything downstream of that — identity, accountability, trust — is suddenly load-bearing in ways nobody designed for.

That's not a payments problem. It's a primitives problem. And I haven't been able to think about much else since.

I think this is the most interesting unseized problem at the intersection of Web3 and consumer AI. I'm building a thesis.

If you're thinking about the same thing, I want to talk.