Kevin's avatar
Kevin
Part time degen

Ethereum blocks can now be proved in real time. Proving them safe is the next job.

Oct 22, 2026. About ZK, Ethereum, and node costs.

On October 13, 2026, ZKsync announced Airbender Prime, a prover that runs on one NVIDIA RTX PRO 6000 Blackwell GPU. ZKsync tested it on 7,200 Ethereum mainnet blocks. It proved 99% of them in under 10 seconds and every block within 12 seconds, which is one Ethereum slot (Crypto Briefing).

Two days later, Vitalik Buterin called the speed battle "definitively won". He also said the work now moves to security. He wants formal verification that a prover matches the Ethereum spec, and he wants limits on what proving can cost relative to gas.

I think he is right, and the reason is worth a few minutes of your time if you hold ETH or run a node.

Why fast proofs matter to every node

Today every full node checks every block by running every transaction in it again. That cost grows with the gas in the block. Each node also has to keep the full execution state on disk. So when Ethereum raises the gas limit, it also raises the hardware you need to run a node.

A zkEVM proof changes that. The node downloads a small proof and checks it. It does not need the execution state. The Ethereum Foundation says a 60M-gas block and a 600M-gas block take about the same time to verify (EF zkEVM blog). That is the property that lets the gas limit rise while a cheap machine can still verify the chain.

The catch was always time. A proof that arrives a minute after the block is useless to a validator that must attest within the slot. Real-time proving means the proof is ready before the attestation deadline.

The speed target is met by more than one team

The Ethereum Foundation set the bar at 99% of blocks proved in under 10 seconds on target hardware. Two teams have published results against it.

The move from 16 GPUs to one is the part I care about. One workstation GPU is hardware that a solo staker or a small operator can buy. Proving stops being something that only a data center can do. You can compare provers yourself on the ETHproofs leaderboard.

EIP-8025 puts proofs on mainnet without making them required

The protocol side is EIP-8025, Optional Execution Proofs. The EF zkEVM team proposed it for Hegotá, the fork planned after Glamsterdam. It adds two opt-in roles to the consensus layer.

The EVM does not change. A validator that does not opt in sees no change. In this first phase, re-execution still decides attestations, and a proof is an extra check. If a node does not get enough valid proofs in time, it attests the way it does today.

That design is cautious, and I think the caution is correct. A bug in a prover in this phase costs some rewards. The same bug in a mandatory design could split the chain. Making proofs mandatory needs a later EIP and a later fork.

The open problems are now about safety and worst cases

Average speed is solved. Three problems are still open.

  1. Correctness. The prover must match the Ethereum spec exactly. Buterin asks for formal verification, which means a mathematical proof that the code does what the spec says. Tests alone are not enough.
  2. Worst-case blocks. The EF notes that single-digit-second proving on average helps little if an attacker can build a block that takes minutes to prove. EIP-8200 would replace some low-usage precompiles with EVM bytecode to cut that worst case.
  3. Cost. If a block can cost far more to prove than it pays in gas, provers lose money on it. Buterin wants a bound on prover cost relative to gas.

The work is moving. At the L1-zkEVM breakout call on October 14, 2026, the Lighthouse team said they had merged support for a proof engine sidecar and can run a node without an execution client. Teams also reported Sepolia testing plans and work on RISC-V compliance tests.

What to watch next

If you follow this topic, these items tell you more than any new speed record.

My view: real-time proving is the most important Ethereum scaling result this year. It makes a higher gas limit possible without pushing home validators out. The headline numbers came first. The value depends on the slower security work that follows. Nothing here is financial advice.

Sources