安全性显然是 L1 的护城河，那隐私呢？我认为，隐私确实会强化通用型（VM）公链的护城河，但并不会动摇 L2 应用链（appchain）这一长期架构趋势。 当下，运行一个应用最具扩展性、也最可定制的方式，是让它成为一条独立的 L2 链，拥有自己的服务器（排序器 / sequencers），并将交易数据批量提交到一个安全、高性能的 L1 共识协议上。 目前全球已有数百万个 Web 应用，仅金融科技（FinTech）应用就有数以万计。因此，我们理应预期未来会出现大量 L2 链，而它们最终只由极少数 L1来保障安全——例如 Espresso，其核心作用只是最终确认交易数据序列的排序。 正如 @alive_eth 所指出的那样，将整个应用从一条支持隐私的通用 VM 链迁移到另一条链，在完全保持隐私的前提下，确实可能非常困难。但这并不意味着：应用链之间实现隐私保护下的互操作是困难的。 为说明这一点，我们可以考虑一种基于 TEE（可信执行环境）的隐私链最简单实现方式： 每条链都是一系列交易 这些交易都使用某个特定 TEE enclave 程序的密钥进行加密 在这种模型下，互操作是直接且清晰的。例如： 链 A 上的 TEE 处理一笔交易 该交易生成一条发送给链 B 的消息 比如：在链 A 上销毁 10 枚 USDC 触发一条在链 B 上铸造 10 枚 USDC 的指令 这条消息会使用链 B 的 TEE 密钥加密，并由中继者（relayer）发送至链 B 的 TEE enclave。外部观察者只能看到两条链在通信，但无法得知任何交易内容。 同样的思想，也适用于完全基于密码学的实现方式（即不依赖 TEE），只是实现机制会更加复杂。 一个关键但容易被忽视的限制 真正的问题不在于跨链或互操作本身，而在于：匿名集（anonymity set）的碎片化。 简单来说：链 A 上的加密交易只在“链 A 的交易集合”内部是不可区分的 如果众所周知只有 John Doe 在使用链 A，那么即使交易内容被完全隐藏，观察者仍然可以判断：John 什么时候在进行交易。 这种元数据泄露问题，可以通过密码学方式缓解：例如，让所有发布到 L1 的交易在外部看来完全一致，只有链 A 的 TEE enclave（或更广义地说，负责链 A 状态转换的运营方）才能私下取回链 A 的真实交易数据。 但这会阻碍主流采用吗？未必。事实上，Web2 早已运行在一个极为相似的隐私模型之下： ISP 或被动的网络监听者 可以轻易区分 PayPal 流量和 Zelle 流量 而 Web2 的整体隐私模型其实更弱：用户只要求对公众隐私。并不期待对应用服务器本身隐私 例如 PayPal、Zelle 在合规要求下，必须看到完整交易细节 总结 我认为，隐私将是加密技术实现主流采用的极其关键特性。 隐私也可能成为通用型 VM 链的重要护城河，因为它会显著提高整个应用迁移的切换成本。 但与此同时，这些因素并不会改变一个更根本的趋势： 👉 每个应用运行在自己独立的 L2 链上，依然是最现实、最具扩展性的架构选择。 这一模式与 Web2 的运行方式高度相似。事实上，今天那些拥有数亿用户的 Web2 应用，完全有可能逐步迁移为 Web3 的 L2 链。而我预计，在这一转型过程中，这种架构将成为主导范式。 Security is clearly a moat for L1s, is privacy? I believe privacy strengthens moats for general-purpose (VM) chains, but does not break the L2 appchain thesis. The most scalable and customizable way to run an application today is as its own L2 chain with its own servers (sequencers) batch-posting transaction data to a secure, high-performance L1 consensus protocol. There are millions of web apps today, including tens of thousands of fintech apps alone. We should therefore expect to see many L2 chains, all secured by at most a small number of L1s like Espresso whose primary role is simply to finalize the ordering of transaction data sequences. It may be true, as @alive_eth argues, that migrating an entire application from one privacy-preserving general-purpose VM chain to another could be difficult to do in a fully privacy-preserving way. However, privacy-preserving interoperability between appchains is not inherently hard. To illustrate this, consider the simplest implementation of privacy chains based on TEEs. Each chain is a sequence of transactions encrypted under the key of a specific TEE enclave program. Interoperability in this setting is straightforward. A transaction processed by the TEE for chain A generates a message destined for chain B, for example burning 10 USDC on chain A triggers a message to mint 10 USDC on chain B. The message is encrypted under the TEE key for chain B, and a relayer delivers it to the TEE enclave for chain B. Observers learn nothing beyond the fact that the two chains are communicating. The same idea applies to purely cryptographic implementations that replace TEEs, though the mechanisms are more complex. There is one important caveat. The main issue is not bridging or interoperability, but fragmentation of anonymity sets. Put simply, encrypted transactions on chain A are only indistinguishable from other transactions on chain A. If it is widely known that only John Doe uses chain A, then even if transaction contents are hidden, observers can still see when John is transacting. This form of metadata leakage could be mitigated using cryptographic techniques whereby all transactions published to the L1 appear identical to public observers and only the TEE enclave for chain A, or more generally the operators responsible for computing chain A state transitions, can privately retrieve chain A’s transactions. But is this actually a problem for mainstream adoption? Web2 already operates under a very similar model. An ISP or passive internet observer can distinguish PayPal traffic from Zelle traffic. In fact, Web2 generally has a weaker privacy model overall, where users require privacy from the general public but do not expect privacy from the application servers themselves, such as PayPal or Zelle. In many cases, regulations explicitly require those servers to see transaction details. In summary, I believe privacy will be an extremely important feature for mainstream crypto adoption. Privacy could also become a meaningful moat for a general-purpose VM chain by increasing the switching costs of migrating an entire application. However, these dynamics do not interfere with the natural tendency toward each application running as its own L2 chain. This remains the most practical and scalable architecture, and it closely mirrors how Web2 operates today. In fact, today’s Web2 applications with hundreds of millions of users can plausibly transition to operating as Web3 L2 chains, and I expect this model to dominate as that transition occurs.