Staghunt and the Missing Layer of the Internet

The internet solved connection. It never solved coordination. That gap — the difference between being able to talk and being able to act together — is what Yonatan Sompolinsky, the founder of Kaspa and one of the researchers behind the GHOST-GHOSTDAG-DAGKnight family of proof-of-work protocols, is now betting is worth building infrastructure for.

His new venture, Staghunt, introduced in a recent Oxford Union address and expanded in a six-page manifesto, argues that crypto's first era gave individuals sovereignty — self-custody, privacy, censorship resistance, permissionless rails — but never built the group version. People can exit a system on their own. But they cannot reliably say, *"I will act, but only if enough of you act with me."* That missing primitive, Sompolinsky contends, is the foundation of an entirely new market category: coordination markets. source

The Internet's Commitment Problem

The Staghunt thesis begins with a deceptively simple reframing. Sompolinsky describes the internet not as a finished product but as a young democracy — one with free speech, constant participation, and global reach, but almost no institutions that let people make credible commitments to each other.

Communication, by itself, does not bind anyone to action. People can form group chats, discover shared preferences, and build entire communities around common grievances. But talking does not create movement. And so obvious collective moves routinely fail to happen.

The manifesto offers concrete illustrations. A community may want to leave a platform everyone resents, but nobody wants to leave first. Capital may want to migrate from an inferior venue to a better one, but no liquidity provider wants to land in an empty pool. A group may share the same conviction, but each member is waiting for proof that enough others will actually show up.

This is not a niche problem. It is the default state of the internet — rich in speech, weak in commitment.

Azazel, Not Moloch

Coordination failure has a well-known villain in crypto and technology circles: Moloch, the demon of misaligned incentives. The usual diagnosis starts with the Prisoner's Dilemma — people defect because individual incentives push them away from collective good.

Staghunt argues this diagnosis is incomplete.

"The demon was never defection. People want to cooperate. They just can't afford to be first and exposed. That's Azazel, not Moloch," Sompolinsky said.

The distinction matters because it changes the solution. If the problem is selfishness, the answer is stronger institutions, stricter rules, or central planners. But if people already *want* to cooperate and simply cannot afford to move first, then the problem is infrastructural. The missing ingredient is assurance — a mechanism for people who already want the same outcome to move toward it together, without exposing themselves prematurely.

Sompolinsky illustrates this with a deliberately human example. A tenured professor might come out of the closet provided five colleagues do so. An assistant professor might require fifty. Neither is unwilling to act. Both simply need assurance that they will not be left standing alone. Under higher stakes — organizing dissent where moving first could mean imprisonment, or platform migration where millions are trapped by the cost of leaving first — the same logic holds at scale.

The Intendo: A New Coordination Primitive

Staghunt's answer is a concept called an intendo: a conditional commitment that only executes once predefined conditions are met. Instead of saying, "I will do this," a participant says, "I will do this if enough others do it too."

This may sound like a pledge, but Staghunt's thesis depends on properties that ordinary pledges lack. The first is coordinated atomicity: if a "pack" forms, the relevant commitments must fire together or not at all. A half-executed pack would recreate the exact failure Staghunt is designed to solve — one person's commitment triggers while others fail to move. The second property, as outlined in the manifesto, is accumulation opacity: participants, operators, and outside observers cannot see the full state of commitments until conditions are met, preventing the signaling and exposure problems that currently make collective action dangerous.

Together, these properties create something the internet has never had: a way for groups to commit to shared action without anyone needing to reveal their position prematurely or risk being stranded.

Beyond One Chain: A New Market Category

Staghunt is not positioning itself as another blockchain project. While Sompolinsky is best known as the founder of Kaspa, Staghunt is described as an independent vision built on top of the network, exploring what coordination markets could look like in practice while following Kaspa's decentralized ethos.

The venture's framing is deliberately broader than crypto. If Bitcoin solved stateless money and Ethereum explored programmable money, coordination markets may represent an equally large unexplored category: infrastructure for collective action. Crypto is Staghunt's first market, not its category. The category is coordination itself.

This framing has implications well beyond cryptocurrency. The question for investors and builders is whether crypto's next decade will be defined by new financial products — or by entirely new ways for groups of people, capital, and eventually AI agents to organize themselves.

That last point deserves attention. Sompolinsky's manifesto explicitly identifies AI agents as future participants in coordination markets. Not as tools humans use to coordinate, but as autonomous entities that will need infrastructure to organize themselves — to pool resources, make conditional commitments, and act collectively without centralized orchestration.

What This Means for Teams Building on Self-Hosted AI

The idea of coordinated AI agents is not theoretical for teams already running multi-agent systems. When you deploy several specialized AI agents — a researcher, a coder, a designer, a project coordinator — on your own infrastructure, you are already facing a coordination problem. Which agent acts first? How do they share context? What happens when one agent's output is a precondition for another's input?

The teams furthest ahead in practical AI deployment are not using single chatbots. They are running orchestrated groups of agents that need to coordinate work across tasks, maintain shared memory, and avoid duplicating effort. The primitives Staghunt describes — conditional commitments, coordinated atomicity, opacity until action — map directly onto the problems multi-agent systems solve daily, even if the terminology differs.

The broader takeaway is that coordination is not just a crypto problem or a governance problem. It is the core engineering challenge of any system where multiple actors — human, artificial, or mixed — need to move together without a central coordinator dictating every step. Whether Staghunt's specific primitives catch on or not, the diagnosis feels right: the internet connected billions of people and machines, but the infrastructure for *moving them together* is still missing.

The Open Questions

Staghunt is early-stage. The manifesto and Oxford Union address lay out a thesis and a vocabulary — intendo, coordinated atomicity, accumulation opacity, coordination markets — but the source material does not detail a live product, a timeline, or tokenomics. How these primitives will be implemented at protocol level, what Kaspa's role will be in practice, and whether coordination markets can achieve the scale and security assumptions Sompolinsky envisions remain open questions.

What is not in question is the gap itself. Across crypto, across distributed teams, across the growing deployment of AI agent swarms, the need for reliable group coordination without centralized trust is clear and growing. Sompolinsky is not inventing the problem. He is naming it and proposing infrastructure to address it.

For builders working at the intersection of decentralized systems and AI, Staghunt is worth watching — not as a finished product, but as a signal that the coordination layer of the internet is finally getting serious attention. The teams that figure out how to make groups act together — whether those groups are made of people, agents, or both — will define the next era of infrastructure.

---

*Based on CoinDesk's coverage of Staghunt, published June 30, 2026. Read the original.*