At the ETHSF Verifiable Computation: From Zero-Knowledge to Infinite Intelligence, Michal Zajac, Head of Research at Nethermind, offered a compelling vision for how cryptography and Ethereum will converge to drive the next wave of innovation in artificial intelligence and user autonomy. 

His central thesis: blockchains aren’t just infrastructure for decentralized finance — they’re becoming the native habitat for intelligent agents.

This isn’t a distant future. As Zajac reminded the audience, “We should stop saying years. Smart contract agents are already here — and governance agents may follow in weeks."

The event was co-hosted by Nexus and Nethermind, sponsored by Halliday, and in partnership with Blockchain Builders Fund and ETHSF.

The rise of new users: From humans to agents

Today’s blockchain users are humans — organic, mistake-prone, and reliant on UX. But tomorrow’s users, Zajac said, will be software agents: digitally-native, protocol-fluent, and ultimately autonomous. These agents won’t care about interface friction or wallet prefixes. They’ll speak directly in protocols and act onchain as first-class participants in economic, governance, and coordination layers.

But autonomy isn’t just a technical feature — it’s a question of trust. And that brings us to verifiability.

From trust to verifiability: How agents earn autonomy

For an agent to act independently on behalf of others (or itself), its inputs and actions must be provable. Zajac outlined three essential pillars of agentic verifiability:

1. Verifying setup — Trust must be rooted in cryptoeconomic guarantees and decentralized infrastructure.
2. Verifying data — Data provenance becomes critical, especially as models ingest real-world inputs.
3. Verifying execution — Inference integrity is crucial, but most current approaches (e.g. zkML, TEEs, fingerprinting) only verify the last step in a multi-stage training and inference pipeline.

As he noted, “We often talk about inference verifiability — but overlook the messy, multi-step process that comes before. Without visibility into data collection, training, and tuning, verifiability is incomplete.”

Protocol, governance, culture: A three-layer blockchain stack

Zajac’s framing of blockchain evolution added philosophical weight to the talk. He described Ethereum as having three intertwined layers:

• Protocol — execution, consensus, fork choice
• Governance — human-led client maintenance, EIP processes, and research forums
• Culture — the diffuse values and norms that shape what the community considers acceptable or urgent

Agents may soon be active at all three layers, not just the protocol. Imagine agents participating in governance forums, submitting EIPs, or even contributing to consensus maintenance. But culture — the informal, human heart of blockchains — may be the last layer agents penetrate. 

Redesigning cryptoeconomics for a plurality of agents

Current cryptoeconomic systems assume rational, money-maximizing actors. That model works well for staking and consensus today — but it’s not built for a world of diverse agents with diverse utility functions.

Some agents will maximize profits. Others might optimize for public goods, reputation, or model performance. Zajac pointed out that economics already has a long-standing framework for this tension: Arrow’s Impossibility Theorem, which shows how competing preferences and constraints create tradeoffs in collective decision-making.

“We’re entering a world where not all actors are economically rational. That’s not a bug — it’s the feature that makes agent ecosystems powerful.”

Towards a more useful proof-of-work

Finally, Zajac sketched a future for blockchain consensus beyond traditional proof-of-stake or legacy proof-of-work. He called for a new direction: Proofs of Useful Work, such as zero-knowledge proofs for AI model training, inference, or data provisioning.

Instead of burning cycles on hashes, nodes could help train models, verify inferences, or supply trusted datasets. This vision could align incentives between blockchain participants and the broader AI ecosystem, driving collective utility while preserving decentralization.

Looking ahead: Agents as first-class blockchain citizens

Zajac’s talk made one thing clear: autonomous agents are not a side quest — they are the next era of blockchain evolution. Ethereum and similar systems are laying the groundwork for agent-native infrastructure, where trust is programmable, governance is pluralistic, and value creation is multidimensional.

As the landscape shifts from humans to agents, from transactions to coordination, and from incentives to intentions, we’ll need new models of design, verification, and consensus. Blockchains — and the cryptographic primitives they’re built on — are uniquely suited to meet that challenge.

“Humans may remain the culture. But everything else — protocol, governance, and economy — will soon have agents in the loop.”