Ethereum co-founder Vitalik Buterin has unveiled a plan for two key changes to the network's execution layer: a shift to a binary state tree and a long-term replacement of the EVM.
Now, execution layer changes. I've already talked about account abstraction, multidimensional gas, BALs, and ZK-EVMs.
— vitalik.eth (@VitalikButerin) March 1, 2026
I've also talked here about a short-term EVM upgrade that I think will be super-valuable: a vectorized math precompile (basically, do 32-bit or potentially…
“These are ‘deep’ changes that many shy away from, considering an incremental approach more ‘pragmatic’,” he noted.
According to Buterin, the state tree and virtual machine account for over 80% of the network's technical limitations. As a result, they become "practically mandatory" for client scenarios utilizing proofs.
New State Tree
The state tree modification is based on the EIP-7864 proposal. It suggests transitioning from the current hexadecimal Merkle-Patricia tree to a binary tree with a more efficient hash function.
The binary structure will reduce Merkle branches by four times, decreasing data volume for lightweight clients like Helios. Switching the hash function to Blake3 or one of the Poseidon variants will further enhance proof efficiency by three to 100 times. However, Buterin acknowledged that Poseidon requires additional security auditing.
This proposal from the Ethereum co-founder is an evolution of earlier plans to implement Verkle Trees. Interest in binary structures was revived in mid-2024 due to concerns that current cryptographic methods might be vulnerable to quantum computers.
New Virtual Machine
Buterin also revisited the idea of replacing the EVM with RISC-V—an open instruction set architecture already utilized by most ZK provers. He outlined a three-phase plan:
- Implement RISC-V exclusively for precompiled contracts.
- Allow users to deploy contracts on the new architecture.
- Completely phase out the old EVM, transforming it into a smart contract operating on top of the new virtual machine.
“The essence of Ethereum lies in its versatility. If the EVM is not good enough for its implementation, we must address the issue directly and create a better virtual machine,” the programmer stated.
However, this initiative has faced resistance. In November 2025, researchers from Offchain Labs (developers of Arbitrum) argued that WebAssembly (WASM) is better suited for long-term development.
Their main argument is that while RISC-V excels at ZK proofs, the "interface architecture" and "proof architecture" do not have to align.
AI
Buterin also suggested that the implementation of the Ethereum roadmap could accelerate thanks to vibe-coding.
This is quite an impressive experiment. Vibe-coding the entire 2030 roadmap within weeks.
— vitalik.eth (@VitalikButerin) February 28, 2026
Obviously such a thing built in two weeks without even having the EIPs has massive caveats: almost certainly lots of critical bugs, and probably in some cases "stub" versions of a thing… https://t.co/ZlTg0r2hvI
He noted that AI "significantly speeds up coding." The community should be open to the possibility that project plans could be completed "much faster and with a higher level of security than expected."
This comment came after an experiment where a developer created a reference implementation of the roadmap in just a few weeks.
Two weeks ago I made a bet with @VitalikButerin that one person could agentic-code an @ethereum client targeting 2030+ roadmap. So I built ETH2030 (https://t.co/2k83PyUP4z | https://t.co/P0A6aHDZBX).
— YQ (@yq_acc) February 24, 2026
702K lines of Go. 65 roadmap items. Syncs with mainnet. Here's what I found. https://t.co/i6gkqLc82L
However, Buterin pointed out the "massive caveats": the high speed of coding "almost guarantees critical bugs," and in some cases, AI may have only created "stubs" instead of working versions.
“But six months ago, even such results were unattainable, so now the trend direction is what matters most,” he added.
The programmer believes the optimal approach is to use neural networks to enhance security, not just to speed up processes. Among potential solutions, he highlighted generating additional test scenarios, applying formal verification, and creating alternative implementations of the same components.
Recall that in February, Buterin discussed the potential process of merging Ethereum with artificial intelligence.