V God announces: EVM will be retired soon, RISC-V will be deployed in three phases, and AI will accelerate Ethereum's evolution beyond expectations

Ethereum Co-Founder Vitalik Buterin recently posted a series of articles outlining two major core transformations for the Ethereum execution layer—changing the state tree to a binary tree via EIP-7864 and replacing EVM with RISC-V in the long term. He pointed out that these two changes account for over 80% of the proof efficiency bottleneck. More notably, Vitalik also expressed that AI will significantly accelerate the entire roadmap and suggested investing half of the speed benefits brought by AI into security enhancements.
(Background: EVM should step aside! Vitalik proposes switching to RISC-V, which can boost ZK proof performance by over 100 times)
(Additional context: Technical debt is mounting; Ethereum chooses RISC-V to “start over”)

Table of Contents

Toggle

• Binary Tree: A One-Time Integration of a Decade’s Experience
• RISC-V Replacing EVM: Making Ethereum “Beautiful Again”
• AI Acceleration Roadmap: “Completion Speed Will Far Exceed Expectations”

Recently, Vitalik Buterin published a lengthy post on Farcaster, detailing his full vision for future upgrades of Ethereum’s execution layer. He admits that these two changes—replacing the state tree and virtual machine—are deep modifications that many instinctively avoid, but because they account for over 80% of the efficiency bottleneck, “gradual fixes” are no longer sufficient.

Binary Tree: A One-Time Integration of a Decade’s Experience

The first change is driven by developers like @gballet through EIP-7864, which replaces Ethereum’s current 16-branch Keccak Merkle Patricia Tree (MPT) with a binary tree structure, paired with more efficient hash functions.

Vitalik listed the core advantages of the binary tree:

• Merkle branch length reduced by 4x: The branch length in a binary tree is 32×log(n), much shorter than the 512×log(n)/4 in a 16-branch tree, directly reducing data bandwidth requirements for light clients like Helios by 75%
• Proof efficiency skyrockets: Shorter branches alone bring 3-4x improvements; switching to Poseidon hash functions could potentially boost efficiency by 100x
• Significant gas cost reduction: Binary trees group storage slots into “pages” (64-256 slots), with block headers and the first 1-4 KB of code and storage sharing the same page. Many DApps load large amounts of data from just a few storage slots, saving over 10,000 Gas per transaction
• Client ZK proofs: If developers want ZK applications to directly integrate with Ethereum state rather than building separate trees, the state tree itself must be proof-friendly

Vitalik describes the binary tree as an “omnibus integration”—a one-time implementation of all the research and insights accumulated over Ethereum’s past decade regarding ideal state trees.

RISC-V Replacing EVM: Making Ethereum “Beautiful Again”

The second change is more long-term and still in the non-consensus phase: replacing EVM with RISC-V. Vitalik mentions, “The protocol has become increasingly ugly and full of special cases over time because people have a subconscious fear of ‘using EVM’.” He finds this unfortunate—Ethereum’s core value lies in its versatility, and if EVM isn’t good enough, it should be addressed directly rather than circumvented.

Advantages of RISC-V include: native execution efficiency high enough to make most precompiled contracts unnecessary; currently, ZK provers are already written in RISC-V, so adopting it directly removes an entire translation layer; a RISC-V interpreter only needs a few hundred lines of code, embodying what a blockchain virtual machine should feel like.

Vitalik proposed a three-phase incremental rollout plan:

• Phase 1: New VM used only for precompiled contracts—80% of existing precompiles will be converted to RISC-V code
• Phase 2: Users can deploy RISC-V smart contracts directly
• Phase 3: EVM is officially retired, becoming a “smart contract written in RISC-V”

Existing EVM users will enjoy full backward compatibility; the only change will be gas cost adjustments—these will be addressed through future scaling efforts over the coming years.

AI Acceleration Roadmap: “Completion Speed Will Far Exceed Expectations”

Beyond the technical roadmap itself, Vitalik expressed high optimism about AI’s potential in Ethereum development. He pointed out that AI could accelerate the entire roadmap and raise security standards.

He highlighted an experiment: someone used “vibe coding” (natural language instructions to AI for code generation) to complete the entire 2030 Ethereum roadmap in just a few weeks. Vitalik was stunned:

“Six months ago, this was far beyond the realm of possibility, and the real question is which direction the trend is heading.”

He also personally tested AI’s coding ability—building a system similar to his blog software within an hour on a laptop, demonstrating the practical efficiency of AI-assisted development.

However, he does not advocate for speed at all costs. Vitalik suggests splitting the speed benefits of AI equally: half to accelerate the roadmap, and half to enhance security—including adding more test cases, formal verification, and implementing independent client implementations to ensure cross-verification of code.

He concluded: “People should remain open-minded about this possibility—not certainty, but possibility—that the speed of Ethereum’s roadmap could far surpass expectations, and security standards could be much higher than anticipated.”

Notably, Vitalik also mentioned a short-term EVM upgrade in the same series: vectorized math precompile, likened to “EVM’s GPU.” This upgrade would enable simultaneous 32-bit or 64-bit operations on numeric lists, potentially accelerating hashing, STARK verification, homomorphic encryption, and other operations by 8-64x.

From binary trees to RISC-V, and then to AI-accelerated development—Vitalik’s vision is not just a technical roadmap but a new development paradigm of “doing more in less time, and doing it more securely.” For the Ethereum community, this is both a source of confidence and a direct challenge to gradualism.