Vitalik Buterin: GKR will become the "highway" of ZK technology.

Ethereum co-founder Vitalik Buterin published a new technical article “A GKR Tutorial” detailing a new proof protocol called “GKR (Goldreich–Kahan–Rothblum)”. He pointed out that this technology is the key driving force behind today’s “ultra-fast ZK proofs”. Although this article is filled with mathematical derivations, the core message is very clear: GKR transforms zk-SNARKs from theory to practice, making complex on-chain computations and AI verification feasible, immediate, and low-cost.

zk-SNARKs (ZK) enters the “acceleration” era

Vitalik pointed out at the beginning that the crypto world is entering a new phase of “ZK acceleration.” From the instant verification of Ethereum L1 on dozens of consumer-grade GPUs to laptops performing two million hash operations per second, and even large language models (LLMs) being able to verify inference processes through ZK technology — these seemingly cross-domain breakthroughs actually have a common foundation: the GKR protocol.

Why GKR is needed: Making massive computations “trusted” but not “burdensome”

Traditional ZK systems can verify the correctness of results, but at a high cost: each layer of computation requires the establishment of cryptographic commitments, leading to a dramatic increase in computational costs.

Vitalik pointed out that GKR changed this model. It only needs to commit to the “inputs and outputs” and can skip the intermediate layer while still ensuring the overall computation is trustworthy. This means that:

Speed increased several times

Significant decrease in costs

Validators only need to check the beginning and the end to trust the entire process.

He described this structure as “designed for multi-layered, repetitive computations,” such as large-scale hashing or AI neural network inference.

From Crypto to AI: GKR's Multiple Application Scenarios

Vitalik uses the Poseidon hash function as an example to demonstrate how GKR handles layered computational structures. However, the potential of this technology goes far beyond a single application.

GKR can be applied to:

1. ZK-EVM verification: Accelerate the operation of the Ethereum mainnet layer (L1) to achieve instant on-chain proof; 2. AI model verification (ZK-ML): Ensure that AI inference results can be cryptographically verified; 3. Batch encryption and data verification: Improve efficiency in high-repetition scenarios such as transaction verification and batch signing.

Vitalik emphasized: “As long as the computation can be expressed as a multi-layer structure, and each layer can be simplified to a low-degree polynomial, GKR can be directly applied.”

Costs have significantly decreased, but there are still challenges.

According to Vitalik's tests, GKR theoretically has only about 15 times the computational overhead, significantly reduced compared to the traditional STARK's 100 times. After implementation, it can even be reduced to within 10 times; if the batch structure is further optimized, the overhead is almost close to 0.

Vitalik also reminded that while pursuing speed, attention must be paid to protocol security. If designed improperly, malicious actors may predict the Fiat–Shamir random challenge values, thus manipulating the proof. For developers, this represents a new design challenge after the ZK system enters the engineering phase.

If we were to use a metaphor to describe GKR, Vitalik believes it is like “a highway built for the world of zero-knowledge” — it does not change the destination (proof of correctness), but allows data to pass through more quickly and lightly. In an era where blockchain and AI technologies are accelerating their integration, this “highway” is not only about efficiency but also the core infrastructure driving the trust foundation of Web3.

This article Vitalik Buterin: GKR will become the “highway” of ZK technology first appeared on Chain News ABMedia.