Vitalik’s Hong Kong Dialogue: “I Hope Everyone Will Do Something Completely Different from Past Ethereum”
TechFlow Selected TechFlow Selected
Vitalik’s Hong Kong Dialogue: “I Hope Everyone Will Do Something Completely Different from Past Ethereum”
Vitalik revealed that Ethereum’s L1 computational scalability target is a 1,000x improvement, and the team has now entered the practical攻坚 phase.
Navigating Web3 tides with focused insightsCompiled & Translated by TechFlow
Guest: Vitalik Buterin (Co-founder of Ethereum)
Host: Qiu Dagen
Podcast Source: Wu Shuo Bu Jia Mi (Wu Shuo on Crypto)
Air Date: April 27, 2026
Editor’s Introduction
Vitalik delivered a keynote speech at the opening event of the ETH Hong Kong Community Hub, revealing that Ethereum’s L1 computational scalability target is a 1,000x improvement—and the team has now entered the active implementation phase. He emphasized that L2s should not merely replicate L1 but instead deliver capabilities L1 cannot provide—such as privacy, low latency, and high throughput. On quantum resistance, he outlined a concrete technical roadmap: replacing elliptic-curve signatures with hash-based or lattice-based signatures, then compressing on-chain footprint via STARK aggregation—deployment expected within 3–5 years. Vitalik also issued a stark warning: AI may autonomously discover all client-side DoS vulnerabilities within 2–3 years, potentially reenacting the 2016 Shanghai DevCon pre-conference attack—a top-priority security concern for the Ethereum Foundation.
Key Quotes
The Role and Value of L2s
“A good L2 cannot be a copy of L1. A good L2 should complement L1 by doing things L1 cannot do.”
“Every application will have its own computation, leading to many composability challenges—making inter-application communication extremely difficult.”
L1 Scalability: Data and Computation
“Scaling Ethereum L1 by 1,000x is feasible—but presents many hard problems.”
“A scalable ERC-20 token will differ significantly from today’s ERC-20.”
“Many core Ethereum principles cannot fully脱离 L1—so L1 itself must scale substantially.”
Quantum Resistance and Security
“An elliptic-curve signature is 64 bytes; a quantum-resistant one is ~2,000–3,000 bytes—requiring new techniques.”
“Block-producing nodes must generate a STARK aggregating all signatures, proving their validity.”
“If we don’t secure our systems, AI in three years will find every vulnerability—and cause us immense pain.”
Chinese-Speaking Communities and Technological Generations
“When a new technological trend emerges, it creates enormous opportunity for a new generation—because there are no ZK developers with ten years’ experience yet; everyone starts from zero.”
“Reflect on the reasons that first drew you to Ethereum—and ask: what does the world need today? That answer may not even include Ethereum as a technology.”
Chinese-Speaking Communities and Ethereum
Host: It’s deeply moving to reunite after four and a half years. To begin, could you share your personal journey interacting with Chinese-speaking communities since launching Ethereum—and your reflections?
Vitalik:
I first heard about the Chinese community in 2013—before Ethereum existed, when only Bitcoin was around. At the time, others told me that mining farms and exchanges in China were larger than those in the U.S. I found this fascinating—yet English-language media reported nothing about it, so I became intensely curious.
In May 2014, I visited China for the first time—Beijing, Shanghai, Hangzhou, and Shenzhen—meeting numerous local communities, miners, and exchanges. I recall Huobi’s office housing hundreds of employees, while U.S.-based Coinbase and Kraken hadn’t yet reached 100 staff. Many people were already thinking deeply about Bitcoin—and beginning to explore smart contracts.
During my 2015 visit, BitShares was gaining traction. Their approach differed: each application ran its own blockchain, with cross-chain interoperability—but applications remained largely independent. Some brilliant developers built upon BitShares’ codebase.
The first major academic contribution from our community came from developer Qian Youcai, who published a whitepaper introducing RANDAO—a decentralized random number generation scheme. RANDAO later evolved into Ethereum’s mechanism for selecting the next block proposer.
The community was exceptionally vibrant—building diverse applications and launching many ICOs. Around 2020, the pandemic triggered significant shifts. Many seem to have completely forgotten that era—as if it never happened. One major subsequent shift was the rise of ZK (Zero-Knowledge) technology. When a new technological wave emerges, it opens massive opportunities for a new generation—because no one yet has ten years’ ZK experience; everyone starts from scratch. Today, thanks to AI, writing any code has become dramatically easier. Even if you’ve never written Solidity before, you can start today.
Since 2022, with the proliferation of L2s, more people have joined DeFi—and gradually, core development too. Core development remains harder: L2s resemble free markets—if you’re an outsider not embedded in the core community but build excellent software, you can still succeed. Core development isn’t like that. It requires coordinating with existing developers on decisions like which EIPs (Ethereum Improvement Proposals) to adopt, when to raise the gas limit, or when to integrate ZKEVM. This coordination makes participation far more difficult—yet participation is slowly expanding.
Overall, my observation traces a progression—from mining and exchanges, toward deeper, broader engagement across all layers of technology and community. It’s been a rewarding journey—but much ground remains to cover.
L2s Should Not Be L1 Copies
Host: Thank you for that rapid retrospective of the past eleven or twelve years. Since launching Ethereum, you’ve spent considerable time engaging with Chinese-speaking communities—in mainland China, Hong Kong, and elsewhere. Over the past two days, you shared perspectives on L2s, stressing they must offer more than just scalability. How do you assess the current state of L2 development?
Vitalik:
L2s remain critically important—but a good L2 shouldn’t simply replicate L1. Instead, it should complement L1 by delivering capabilities L1 fundamentally cannot. For instance, many applications require blockchain infrastructure—but also need additional features beyond L1’s reach: privacy, high scalability, low latency, or oracles.
The ideal L2 design process begins by identifying applications you believe are truly needed and valuable—then asking where those applications rely on L1, and what common requirements L1 fails to satisfy. This line of reasoning reveals why copying EVM isn’t optimal. Indeed, the most successful L2s today began with EVM compatibility—but their current success stems from incorporating non-EVM components purposefully—not building L2s for L2’s sake.
L1 Scalability: Dual-Track Advancement
Host: You recently published updates on the Ethereum roadmap—I’d especially like to ask about data scalability and computational scalability, both highly anticipated topics. How will scaling proceed? Any directional guidance?
Vitalik:
First, why is L1 scalability important? We observe that many applications—while theoretically deployable on L2s—would require excessive intermediaries if fully offloaded from L1. Each intermediary introduces potential failure points. Thus, Ethereum’s foundational philosophy cannot fully detach from L1—L1 itself must scale substantially.
L1 scalability comprises two dimensions: data and computation—both requiring expansion. Our latest upgrade, Fusaka, introduced PeerDAS—a data-scaling solution. Currently, Ethereum’s chain supports far more data capacity than is being used—only ~25% utilized—but we possess mechanisms to expand capacity tenfold or more, on demand.
Yet data alone isn’t sufficient. Without on-chain computation, anyone could dump arbitrary data onto-chain—but how would users interpret it? What meaning would it hold? If computation remains entirely off-chain, each application develops its own compute layer—leading to severe compatibility and composability issues. Inter-application communication becomes extremely difficult. Hence, Ethereum’s native computation remains indispensable.
We’re now intensively exploring computational scalability. Scaling by 1,000x is technically possible—but fraught with challenges. Application developers will often need to fundamentally rethink their approaches. A scalable ERC-20 token will differ significantly from today’s version. We’re actively researching and prototyping solutions—including using ZK proofs to verify EVM execution per block. These are feasible but require substantial time and rigorous security assurance.
Ethereum’s current complexity demands extreme caution. If compromised, AI in three years will inevitably uncover every vulnerability—inflicting far greater pain. You may recall a complex incident in Shanghai a decade ago: four hours before DevCon began, I was asleep when someone called my hotel room urgently. I rushed downstairs to find Ethereum under a DoS (Denial-of-Service) attack.
We labored for hours to understand and resolve the attack—fixing it roughly three hours before the conference began. We released client PRs (Pull Requests) just in time, and the conference started on schedule—we felt victorious. Yet two days later, a second attack struck. Five days later, a third. We never identified the attacker—but one person, over roughly one month, discovered every DoS vulnerability across Ethereum’s two primary clients. For that entire month, Ethereum’s chain was nearly unusable.
Some users persisted—Augur’s ICO, for example, launched mid-DoS. After patching all known vulnerabilities, stability returned—and few issues recurred. But AI could similarly discover all client vulnerabilities within two to three years—repeating that agony. So we aim to proactively harden clients using novel verification methods—ensuring our code contains no such flaws before AI finds them.
AI and Quantum Threats
Host: Since founding Ethereum, you and your team—and the broader community—have consistently focused on two pillars: scalability and security. Application-layer innovation has largely been community-driven, while you’ve concentrated on scalability and security. Yet you’ve also frequently shared thoughts on AI’s explosive growth and quantum computing. Media coverage abounds with speculation—could you clarify today how you envision AI and quantum computing intersecting with blockchain and Ethereum in the future?
Vitalik:
Both challenges are addressable within Ethereum—but require diligent, sustained effort. Let me illustrate with a non-technical analogy: Imagine a country that has never experienced rain—no one understands the concept. Its buildings lack roofs, gutters, or waterproofing. The first rainfall might destroy 5% of structures. Residents wouldn’t anticipate problems—because rain simply doesn’t exist in their experience.
Then scientists announce: “Rain is real—and will arrive in 5–10 years. Here’s how to build rain-resilient infrastructure.” Everyone understands the theory—but implementation is grueling: inspecting every house, school, office, and public infrastructure for vulnerabilities—and fixing each one individually.
Quantum resistance and AI resilience in Ethereum mirror this challenge. We know precisely what to do within the next 3–5 years. In fact, I wrote my first quantum-resistant signature algorithm five years ago—buried deep in a GitHub repository. Back in 2017, I even deployed a smart contract implementing a hash-based signature scheme. We understand the solutions.
Why haven’t we deployed them yet? Efficiency. An elliptic-curve signature is 64 bytes; a quantum-resistant one runs 2,000–3,000 bytes. Directly embedding such signatures in blocks is prohibitively expensive. Instead, we need aggregation: block-proposing nodes must generate a single STARK proof verifying all signatures’ validity—allowing the original signatures to be discarded, storing only the compact STARK on-chain.
For AI, our current strategy centers on formal verification—to mathematically prove ZK code satisfies security requirements. This is tractable at the L1 layer: there’s only one L1, and it’s finite. Application-layer complexity is far higher—countless apps, each with unique dependencies—making holistic verification vastly harder. Yet this remains a high-priority focus within the Ethereum Foundation.
Host: Would adopting quantum-resistant schemes impact gas fees?
Vitalik:
Yes—by default, they increase costs. Simply replacing elliptic-curve signatures with hash- or lattice-based ones would raise per-transaction gas from ~20,000 to ~200,000—cutting transaction throughput tenfold. Hence, aggregation is essential: block-proposing nodes must replace thousands of individual signatures with a single STARK.
For example: 1,000 users each submit a transaction containing a ~3,000-byte signature—totaling ~3 MB. A node generates one STARK proving those 3 MB signatures’ validity. A well-optimized STARK occupies just 256 KB—or even 128 KB—regardless of signature count. We know the theoretical solutions—but translating theory into production-grade implementation remains the critical next step.
Message to Hong Kong
Host: Your return is warmly welcomed—especially by Chinese-speaking communities, thrilled to see you back regularly. Finally, we’d love your message to Chinese-speaking contributors building Ethereum, driving community growth as builders, and shaping future development. Also, today marks the official opening of the ETH Hong Kong Community Hub—what hopes do you hold for this Hub?
Vitalik:
The new Ethereum ecosystem offers countless opportunities—and recent years present a rare chance to rethink fundamentals, empowered by mature ZK technology and AI. AI has slashed code-writing costs tenfold—now enabling ordinary people to build meaningful tools.
Even as new tools help us meet emerging demands, those demands themselves grow ever more complex—and the world accelerates relentlessly. I urge everyone to engage from first principles: Don’t start from today’s Ethereum ecosystem or protocol—but from Ethereum’s core philosophy. Reflect on what originally inspired your interest in Ethereum—and ask: What does the world need *today*?
That answer may not involve Ethereum at all. Or it may combine parts of Ethereum, AI-native components, ZK-SNARKs, or secure hardware. Next door lies Shenzhen—home to vast hardware expertise, including open-source hardware and open-source AI. We can begin merging these domains to create entirely new innovations.
I encourage everyone to think deeply—and to build things radically different from Ethereum three years ago.
Host: Returning to our origins: every great project solves real human problems. Thank you, Vitalik.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
深潮TechFlow
