Vitalik Buterin wrote a long post: The GKR cryptographic protocol can quickly prove Ethereum, and zk-ML is accelerating AI LLM.

Vitalik Buterin pointed out GKR, which allows for the immediate generation of Ethereum L1 proofs using a small number of GPUs, opening a new chapter for ZK scalability (Background: Are Ethereum MEV bots money laundering? A US court hears the MIT brothers case for the first time, with sandwich arbitrage of 25 million dollars involved in fraud) (Additional background: Bhutan is migrating the “national digital identification” of 800,000 citizens onto Ethereum, Vitalik: Let the people have peace of mind in controlling their data) Ethereum founder Vitalik Buterin stated on the 19th that the GKR protocol allows zero-knowledge proofs (ZKP) to enter the era of immediate operation: 50 consumer-grade GPUs can handle Ethereum L1, and an ordinary laptop can verify Poseidon hashes 2 million times per second. This breakthrough sets a new benchmark for blockchain scalability and privacy protection. The efficiency of GKR comes from the fact that the GKR protocol is designed for specific computational scenarios, with its core advantage being simplicity. Traditional STARKs require commitments to all intermediate layers of the circuit, often consuming hundreds of FFT and hash functions; GKR only commits to the inputs and outputs, greatly reducing proof size and verification time. Buterin explained: “GKR is a cryptographic protocol designed for this kind of model. Its efficiency comes from avoiding commitments to all intermediate layers.” Taking the Poseidon2 Hash as an example, GKR theoretically reduces the prover's overhead by about fifteen times, while traditional solutions may amplify it by hundreds of times. Efficiency can be further enhanced through Grün's trick, linear batch processing, and partial round optimizations. However, Buterin also reminded developers to ensure that Fiat–Shamir challenges are genuinely random to maintain security. ZK-EVM and zk-ML acceleration High-efficiency provers are immediately evident in ZK-EVM. Brevis's Pico Prism uses a multi-GPU cluster to complete 99.6% of Ethereum L1 block proofs within 12 seconds; ZKSync's Airbender also achieves sub-second verification. These cases demonstrate that GKR-like technologies provide the momentum for real-time block generation. The same structural characteristics make GKR suitable for “multi-layer low-level computations + a large number of repeated functions” in zk-ML. Model inference can be verified without leaking weights and inputs, laying the foundation for AI privacy computing. Wall Street perspective: Efficiency is competitiveness GKR pushes zero-knowledge proofs towards large-scale applications, reminiscent of the performance optimization of TCP/IP in its early days. For investors, high-performance ZK infrastructures and new decentralized applications will become the focus; lower computing costs may also attract mainstream institutions to enter the market. Although current provers still rely heavily on GPUs, if ASICs become available in the future, the economic threshold is expected to continue to drop. GKR solves speed and simplicity, while the privacy layer can be complemented by SNARKs or STARKs. With the maturity of ultra-fast ZK technology, a scalable, efficient, and privacy-friendly Web3 ecosystem is rapidly taking shape. Related reports Ethereum Fusaka upgrade completed Sepolia test, next step is to launch the Hoodi testnet, expected to land on the mainnet in early December The Ethereum Foundation establishes a “Privacy Research Group” to promote six major roadmaps, fully initiating competition in the privacy track A US judge was torched in “physical retaliation”! Vitalik calls: protect judicial personnel with zk voting (Vitalik's long post: GKR cryptographic protocol can quickly prove Ethereum, zk-ML is accelerating AI LLM) This article was first published on BlockTempo “Dynamic Trends in Blockchain News Media.”