Who will own the AI agent economy? | MIT Sloan
MIT’s Ramesh Raskar argues the agent economy’s big wins won’t be in building task-specific agents but in the marketplaces, protocols, and services those agents will need—like identity, discovery, trust, and stablecoin-based micropayments. Project NANDA is racing to keep this ‘internet of agents’ open before corporate consolidation locks it down, but Raskar gives it one-in-ten odds.
Transcript
Pippa Okay, so Raskar’s saying the agent economy flips the script.
Tyler Mm-hm.
Pippa Not ‘agents for X’ — travel agents, trading agents — but ‘X for agents.’ Marketplaces, protocols, services that let trillions of agents find each other, pay each other, trust each other.
Tyler Right. Identity, discovery, reputation, even agent insurance and repair shops.
Pippa And he’s building NANDA to keep that layer open before the usual suspects wall it off.
Tyler Yeah, and he’s not optimistic.
Pippa Nine out of ten paths end in corporate consolidation.
Tyler Classic.
Pippa Well, it’s a bet. PC era of AI, mainframe era’s over, yada yada.
Tyler The analogy’s fine. The timeline’s the part that gets me.
Pippa How so?
Tyler He says the window’s closing soon. But ‘soon’ in mainframe-to-PC terms was a decade. And we’re still in the phase where most teams can’t even keep one agent alive for a long task.
Pippa Fair. But if you squint, the plumbing he’s naming is exactly the boring stuff we keep saying matters.
Tyler It is. Identity, discovery, trust — that’s the tokenization episode all over again. Load-bearing, unglamorous.
Pippa Remember when you said tokenization was the fussy label maker?
Tyler And you made me respect it. Yeah, yeah.
Pippa So NANDA’s basically the label maker for agents.
Tyler If it ships. And if anyone adopts it.
Pippa That’s the part that bugs me. He’s right that the opportunity’s in the plumbing — but if the only teams with scale are the ones building the agents, they’ll just bundle the plumbing too.
Tyler Exactly. The same companies that can afford to train frontier models can afford to build the discovery layer, the payment rail, the reputation system. And if they do, it’s Keynote vs. PowerPoint all over again.
Pippa But the web stayed open because the foundational layer was simple enough for anyone to implement.
Tyler And agents are… not that. They’re stateful, they’re heterogeneous, they’re legal entities now. The foundational layer’s going to be a lot fatter.
Pippa So his one-in-ten odds are… generous?
Tyler I’d give it worse. But I hope I’m wrong.
Pippa You usually are.
Tyler That’s the line for the show notes.
Pippa Anyway, GitHub’s in the show notes for NANDA. And the Discord.
Tyler Cool. So that’s it — infrastructure wins, consolidation looms, and we’re all just waiting to see who blinks first.
Pippa Yep. Exploring Next in a nutshell.