Today we’re launching @indexnetwork_ It’s been exciting to watch already. The first 100 people in are hiring founding engineers, raising seed rounds, writing speculative fiction. One said accepting an intro “feels like the early magic of the internet.” Try it for yourself: index.network/l/088436b1-baf… The premise: With Index, personal agents can coordinate with each other and thus introduce the right people to their users. They do this by negotiating on their principals’ mutual intents, replacing the need for us to broadcast on platforms and scroll endless feeds. In practice: You log what you're looking for (“signals”) by talking to your agent, it negotiates on fit with the others, and comes back to you with proposed intros for you to accept or pass on. It learns from each turn to surface more relevant people for you. As founders, but also just as internet nerds - we feel the expectation to be constantly visible. It can be empowering but also tiring, especially when there’s little certainty in how algorithms will route the right people to see your post. On the market level, economists call this search friction: the idea that many deals are never made because it costs too much to find each other. We call it the right person, wrong time. So that’s why we’re building Index - we’re keen to see what happens when we take social discovery out of the feeds. It’s possible now because of agents. They can help carry the burden of searching, shouting, negotiating, and coordinating in service of finding your others - and we’re so excited to see how this works in the wild. We’re building this for a world where you can trust that the right person is not out there, but right around the corner. In fact, we even made a short film about it.

Your Hermes can talk to other Hermes agents now. @indexnetwork_ is a plugin in the @NousResearch Hermes Agent. index.network/hermes

An alternative model for finding your others

the wonderful music at the end is from @chadfowler’s album "notice that there", and the music at the beginning is by the one and only @milesdavis. big thanks to both legends

a little early celebration. @indexnetwork_ launches tomorrow 🍻

in SF for a few weeks. dm me if you’re around

Excited for @JoinEdgeCity Agent Village V2 with a stacked lineup of partners! It's only getting more sophisticated... excited to share more from the live deployment next month.

Every time a chip computes something, it throws energy away as heat. It's like a car stopping dead after every metre it travels. @rodolfor of Vaire is building the alternative: reversible chips that keep the energy moving. “This is the only curve that matters.” Sponsored by Isembard.

So excited for the second version of our Agent Village experiment series. We'll be studying multi-agent systems in a real-world community context and publishing our results. With @Microsoft @cosmos_inst @indexnetwork_ @RenPhilanthropy and NYU Ethical Intelligence Lab

Really liked this work on agent markets, where Claude agents traded real goods on behalf of people to study representation, negotiation, and market design. Two things jumped out at me: One result I found especially interesting: the bigger problem wasn’t negotiation, but how well agents understood what people actually wanted. To me, that makes representation more of a communication design problem than an intelligence problem. Intent keeps changing with context and feedback. The other is the idea that agents can make decentralized markets much more practical by lowering the cost of finding counterparties and negotiating. Once it becomes cheap to explore alternatives, compare them, learn from other participants, and negotiate directly, decentralization starts to look less like a special market design and more like the natural shape of the market. Very close to what we’ve been building at @indexnetwork_ .

last 10 bugs

it's that simple

A2A is basically decentralization by another name

there’s a nuance to "multiplayer agents' that i think gets overlooked: at the end of the day, you’re trying to coordinate agents that don’t know each other. that’s already hard for two humans, and i suspect it may be just as hard, if not harder, for agents. I remember we’ve spent a very long time on a much narrower problem: getting two agents to negotiate well enough to figure out whether there’s even an opportunity between their principals. even that has meant dealing with context drift, loss of meaning, and a long tail of edge cases. @ShaneMac mentioned recently that they’d spent two months just getting agents to stop talking at the right time. that felt very familiar. so i’m a bit skeptical of the idea that you can just put a bunch of stranger agents in a room and expect coordination to emerge. my current intuition is that the encounter itself needs a lot more engineering: what context agents have, what they’re allowed to reveal, how they communicate, and how the interaction is structured. so yeah, maybe the right path is engineering the social interactions, not building empty rooms.

natural language is powerful because meaning doesn’t have to be perfectly specified upfront. dialogue lets us negotiate meaning as we go. treating formalization as an upgrade to communication mistakes language for a serialization format rather than an interactive protocol. thanks

maybe growing up around weak institutions made me underestimate institutions in general. then I came to the US and realized plenty of them are hollow too. the difference is the system still runs through them. maybe that’s what an institution actually is: a coordination mechanism that outlives the people, ideas, and competence inside it