- Ethereum co-founder Vitalik Buterin said advances in cryptography could end the model in which every blockchain node repeats the same computations.
- Buterin expects SNARKs, STARKs and related technologies to improve scalability and privacy simultaneously.
- He said artificial intelligence could accelerate software development, testing and formal verification for increasingly complex blockchain systems.
Ethereum co-founder Vitalik Buterin said in an online keynote at the 12th Global Blockchain Summit on Sept. 23, 2026, that zero-knowledge proofs, formal verification and artificial intelligence are reshaping blockchain architecture. The shift could improve scalability and privacy without fully abandoning decentralization guarantees, according to Buterin’s address during Shanghai Blockchain International Week.
Buterin said the industry’s technical capabilities and its understanding of the properties decentralized systems should provide have changed over the past 15 years. Early blockchains established who owned assets and who could manage them, while the next stage will involve programmable privacy, in which code also determines who may access specific information.
Early blockchain architecture required many participants to perform the same operations repeatedly. That model supported verifiability but constrained performance and privacy, Buterin said.
SNARKs and STARKs are gradually changing that approach. One participant could perform an operation and generate a cryptographic proof, allowing the remaining participants to verify the result rather than repeat the underlying computation.
“The era when everyone has to repeat everything is coming to an end,” Buterin said.
Recursive STARKs and Ethereum’s architecture
Buterin identified recursive STARKs as a key area of development. The technology can combine multiple proofs into one, reducing the data and computation that must be processed directly on a blockchain.
He previously proposed using that approach in EIP-8288, which envisions aggregating cryptographic signatures and STARK proofs at the mempool level before publishing a compact proof of their validity.
Buterin expects the technology to reduce the cost of private and post-quantum transactions and move some complex cryptographic operations away from Ethereum’s main execution path.
The approach forms part of a broader proposed change to Ethereum’s transaction structure. In early September, Buterin discussed EIP-8141, which divides a transaction into separate actions and dependencies. He said the architecture should allow some operations to run in parallel rather than sequentially.
Buterin also highlighted FOCIL, a mechanism for distributed block building. FOCIL is intended to reduce dependence on a single participant for including transactions in a block and strengthen the network’s resistance to censorship.
In the longer term, Buterin said a blockchain should not be viewed solely as a layer-one network. He described a wider computing pipeline encompassing local operations on users’ devices, mempools with different privacy characteristics, proof aggregation, layer-two networks and the base layer.
Buterin said such an architecture could substantially increase scalability while retaining decentralization guarantees. In the summer of 2026, he described Ethereum as moving toward a notional “third era” focused on scaling, privacy and resilience to quantum computing.
AI and formal verification
Buterin said artificial intelligence could play a distinct role in the transition because modern cryptographic systems are becoming complex enough that manual development and code audits risk becoming a limiting factor.
AI could help write and analyze software, create specifications, run tests and conduct formal verification, he said. Formal verification can mathematically demonstrate that an implementation satisfies its stated requirements.
According to Buterin, combining several independent specifications with automated verification could support significantly more complex systems without a proportional increase in risk. He also linked limitations in current layer-two networks to the difficulty of verifying complex code, saying projects must rely on additional protections, including multiple proof systems and security committees.
Buterin has previously said that pairing AI with formal verification could significantly reshape Ethereum’s security. Algorithms could help identify vulnerabilities, verify protocol implementations and compare code with formal specifications.
He said AI’s potential role extends beyond software development. AI agents could independently interact with infrastructure spanning layer-one and layer-two networks and intermediary services, selecting the appropriate tools for a particular operation.
In February 2026, Buterin presented an updated vision for AI and Ethereum that emphasized decentralization, privacy and verifiability. He urged developers in his latest address to reconsider applications designed when advanced cryptography, programmable privacy and automated formal verification were not regarded as practically viable, saying those constraints are gradually receding.
Source: Incrypted
