At Permissionless IV, which is happening this week in New York City, Nexus Founder and CEO Daniel Marin joined a compelling panel on the future of open and verifiable AI, alongside Illia Polosukhin (NEAR Protocol), Andrej Radonjic (Grass), and Daniel Shapiro (Blockworks Research). Their conversation made one thing clear: the convergence of crypto and AI isn’t about hype — it’s about infrastructure, agency, and creating an open market for intelligence.
A global playing field for AI competition
Daniel Marin opened with a provocation: the real solution to AI centralization isn’t fear — it’s competition. “The solution is acceleration and ensuring there’s maximum competition at every level of the stack,” he said, emphasizing that blockchains can democratize access to capital, data, and compute. “Crypto and blockchains as a financial layer… can drive the best chances that AI evolves in a way that benefits everyone.”
This framing resonated across the panel. Illia Polosukhin underscored the geopolitical edge: while centralized AI models risk surveillance and control, decentralized AI systems offer “confidentiality” and “user-owned intelligence” — a future where agents act on your behalf, not someone else’s.
Economic agency as an AI primitive
Daniel Marin outlined a future where AI systems don’t just perform tasks — they own capital, participate in markets, and improve themselves. “If we close the loop — where AI models can earn money, then use it to improve themselves — we might hyper-accelerate intelligence,” he explained. Nexus is pioneering this model, giving AI agents tools like identity, verifiability, and payment rails.
Daniel Shapiro echoed this shift in framing. While many fixate on decentralizing model training, he argued the real frontier is ownership and deployment. “I find it odd there hasn’t been more push to decentralize recommendation engines — they’ve had a bigger societal impact than most large language models,” he said.
Smart contracts, smarter agents
The panel also explored how crypto tools like smart contracts can underpin autonomous AI agents. Marin asked how agreements between such agents could be enforced — a practical question as agents begin managing resources, users, or even entire businesses.
Polosukhin detailed a multi-layered answer: escrow, auditability, and AI-native arbitration. “The agreement itself has dispute facilities,” he explained, adding that AI can even resolve conflicts unless escalated to traditional courts. These systems are already proving out in NEAR’s ecosystem, where autonomous agents run sentiment-based trading strategies onchain.
Verifiability, not vibes
Throughout, the conversation returned to a shared principle: Verifiability must extend across the stack. From Marin’s vision of economically autonomous agents, to Polosukhin’s call for confidential compute, to Radonjic’s emphasis on distributed data collection, the message was consistent: meaningful decentralization isn’t a vibe — it’s an engineering discipline.
As Marin summarized: “What we should be pushing for is to provide the economic environment… so everyone can innovate.” In a world reshaped by AI, it’s a vision of open competition — not closed platforms — that holds the greatest promise.
The Nexus Verifiable AI Lab is dedicated to exploring the frontier of verifiability, economics, and artificial intelligence in order to expand the boundaries of human and machine cooperation.