During this year’s ETHSF, Nexus Founder and CEO Daniel Marin opened his keynote during the Verifiable Computation: From Zero-Knowledge to Infinite Intelligence event with a challenge to think bigger — not just about blockchain, but about how computation itself is changing. The talk, titled Envisioning a Verifiable World, explored the accelerating frontier of verifiable computation and Nexus’s role in shaping it.

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

From whitepaper to supercomputer

Nexus began with a radical idea: to make verifiability — the ability to prove that a computation happened correctly — a built-in property of computation itself. What started with a whitepaper in January 2024 has now become a fast-evolving succession of zkVM models, each delivering exponential performance leaps.

The recent launch of zkVM 3.0 represents a 1000x speedup over the earliest zkVM model, enabling a “blockchain supercomputer” — a distributed system that accumulates global compute into a single, universal proof.

“Just as OpenAI develops intelligent compute models, we develop verifiable compute models.”

Introducing the Nexus Layer 1

The centerpiece of the keynote was an explanation of the Nexus Layer 1, a minimalist blockchain architecture built for proof scalability. It’s designed to separate execution, consensus, and storage — mirroring traditional computer architecture but applied to an eventually decentralized world computer.

“At the center is a single proof — a proof that concentrates the execution and truthfulness of the whole blockchain.”

In the Nexus view of the world, this architecture unlocks a universal operating system where every user becomes a contributor of compute. The Nexus OS makes this seamless: one-click, mobile-friendly, and designed to blur the lines between users and nodes.

“We’re getting close to a future where we don’t need a supercomputer to prove one block per second. A single MacBook Pro might do.”

This progress signals a near future where proof performance scales with participation — every laptop, every phone, an active node in the global verifiable computer.

Verifiability meets AI

The second half of the talk turned toward the intersection of AI and proof outlining a research frontier where verifiable inference — proving that an AI model was trained and run correctly — is not just desirable but critical.

“How do we prove inference was done correctly? That we’re interacting with the right model? That the training data was valid?”

Here too, the zkVM architecture offers a path forward. Nexus is extending its instruction set, enabling developers to verify AI workloads with familiar tools. The talk closed with a proposal that the convergence of AI and verifiability isn’t just a feature — it’s a singularity.

“AI enhances intelligence. Verifiability enhances trust, financial power, and access. These forces are compounding.”

This is the proof singularity: a point at which intelligence and trust scale exponentially and reshape how we compute, build, and collaborate.

Stay tuned for more ETHSF recap coming soon.