2026.05 Gold, Alloy, and the New Meaning of Peer

[按] 本文是我在 X 上的一篇回应，关于在 AI 时代我们如何鼓励在比特币上的创新。

https://x.com/lorenzonical/status/2050402041915408635

Thank you for sharing this, and congratulations again on making inAI the first real on-chain AI experiment running on Fractal.

I strongly agree with the direction you mentioned: deploying a smarter and heavier model, giving it stronger environment awareness, and exploring the possibility for the model to conduct certain on-chain operations. These are all directions worth pushing further.

From a broader perspective, I have been thinking about one question: in a future where AI agents increasingly participate in on-chain activities, what roles should Bitcoin and Fractal play?

On Ethereum, a smart contract essentially defines a set of rules for how assets can be moved under certain conditions. The process is usually automated, transparent, composable, and relatively immutable after deployment. It allows humans to encode complex financial and application logic directly into on-chain systems, which is one of Ethereum’s most important innovations.

At the same time, complexity itself also introduces new risks. As AI capabilities continue to improve, many smart contract vulnerabilities, architectural weaknesses, and edge cases that previously required top-tier security experts to discover may be exposed by AI in a much more systematic and efficient way. This is somewhat similar to what happened with AI in Go: once AI crossed a certain threshold, even structures designed by the strongest human experts started to reveal previously unseen weaknesses.

From this perspective, Bitcoin’s restraint in functionality is not necessarily a weakness. For a system whose core mission is value storage and value transfer, stability, verifiability, and long-term robustness should always take priority over functional complexity.

If the value of an alloy lies in having more properties and adapting to more use cases, the value of gold lies in having higher purity. Bitcoin is closer to the latter.

This is also why my view on Bitcoin-native programmability has evolved over time. At UniSat, we previously invested significant effort into exploring programmability on Bitcoin, including Stateful Contract, where state can be passed between parent and child transactions through preimage-based information transfer, enabling more advanced composable contract capabilities.

Those explorations were valuable at the time, and they helped us understand the boundaries of Bitcoin Script more deeply. But I think that stage has largely passed. In the coming years, the increased system complexity, expanded security surface, and reduced robustness brought by highly expressive programmability will likely be challenges that Ethereum and other more expressive systems need to continuously face and solve.

For Bitcoin, the most important mission is still to remain a stable, reliable, and long-term trustworthy P2P system for value storage and value transfer.

The meaning of “peer,” however, may evolve.

In the future, a peer will not necessarily only be a human. It may also be an AI agent.

In a world where humans and agents collaborate closely, Bitcoin can become a value storage and payment medium that both humans and AI agents can accept. Fractal’s role is to provide a higher-performance, lower-cost, and more flexible environment for such experiments, while still preserving Bitcoin’s core spirit and engineering consistency.

From UniSat’s perspective, we are currently working on two relatively fundamental but very important things.

The first is packaging wallet capabilities into a command-line tool: UniSat Wallet CLI. This would allow users to let their agents locally store, manage, and send a small portion of their funds, instead of relying entirely on centralized services or browser-based interfaces for every on-chain operation.

The second is upgrading UniSat API into UniSat AI Kit. The goal is to give agents convenient access to Bitcoin protocols and assets, let them query on-chain data and transaction status, and, when necessary, autonomously construct and execute on-chain transactions.

I see these as foundational infrastructure for more advanced AI applications. Only when agents can safely understand on-chain state, manage assets, construct transactions, and execute operations will more truly Bitcoin-native AI use cases naturally emerge.

As for more specific and frontier AI experiment directions, I do not want to define them too narrowly at this stage. One of the most important characteristics of the AI era is that the possibility space has suddenly become larger than ever before. Many truly important directions may not come from the routes that seem most obvious today.

So I would rather see Fractal become an open experimental ground, instead of limiting everyone’s imagination to a few predefined directions from the beginning.

From Fractal’s perspective, we will do our best to support teams that are truly creative, technically capable, and willing to explore Bitcoin-native applications, on-chain intelligence, and agent-driven on-chain operations. Whether through funding, infrastructure, or ecosystem collaboration, if an idea is strong enough, we should find ways to help make it real.

inAI is a very good beginning. What excites me even more is that it may only be the first signal of this class of experiments on Fractal. What is truly worth looking forward to is seeing more teams combine AI, Bitcoin, inscriptions, native assets, and agent operations in ways that may go beyond what we can fully predict today.

(Chinese version follows)

谢谢你的分享，也再次祝贺 inAI 成为 Fractal 上第一个真正跑起来的 on-chain AI 实验。

我也非常认同你提到的方向：更大的模型、更强的环境感知能力，以及让模型具备一定的链上操作能力。这些都是非常值得继续探索的方向。

从更大的视角看，我一直在思考一个问题：在 AI agent 逐渐参与链上活动的未来，Bitcoin 和 Fractal 应该承担什么样的角色。

在 Ethereum 上，智能合约本质上定义了一套“在特定条件下如何转移资产”的规则。这套规则通常是自动化、透明、可组合，并且在部署后具有较强不可变性。它让人类可以把复杂的金融和应用逻辑写进链上系统，这是 Ethereum 很重要的创新。

但与此同时，复杂性本身也会带来新的问题。随着 AI 能力的提升，很多过去只有顶级安全专家才能发现的合约漏洞、架构缺陷和边界条件，未来可能会被 AI 更系统、更高效地暴露出来。这有点像 AI 在围棋中的突破：当能力跨过某个临界点之后，人类顶尖专家过去认为非常稳固的结构，也会开始显现出新的弱点。

从这个角度看，Bitcoin 在功能上的克制，不一定是劣势。对于一个以价值存储和价值转移为核心目标的系统来说，稳定性、可验证性和长期鲁棒性，永远应该比功能复杂度拥有更高的优先级。

如果说合金的价值在于展现更多特性、适应更多场景，那么黄金的价值恰恰在于更高的纯度。Bitcoin 更接近后者。

所以我现在对 Bitcoin-native programmability 的看法也发生了一些变化。我们过去曾经在比特币可编程性上投入过大量精力，包括研究 Stateful Contract，也就是通过 PreImage 在父子交易之间传递状态，从而实现更复杂的组合合约能力。

这些探索在当时非常有价值，也帮助我们更深入地理解了 Bitcoin Script 的边界。但我认为那个阶段已经过去了。未来几年，复杂可编程性带来的系统复杂度上升、安全边界扩大和鲁棒性下降，可能会是 Ethereum 等更强表达能力系统需要持续面对的问题。而 Bitcoin 最重要的任务，仍然是成为一个稳定、可靠、长期可信的 P2P 价值存储和交易系统。

只是这里的 “peer” 的含义会发生变化。

在未来，一个 peer 不一定只是人，也可能是 AI agent。

在人类和 agent 高度协作的世界里，Bitcoin 可以成为人类和 AI agent 都能够接受的价值存储和支付媒介。而 Fractal 的意义，是在保持 Bitcoin 核心精神和工程一致性的前提下，为这些新型实验提供更高性能、更低成本、更灵活的环境。

从 UniSat 的角度，我们现在正在做两件相对基础但我认为非常重要的事情。

第一，是把钱包能力封装成命令行工具，也就是 UniSat Wallet CLI。这样用户可以让自己的 agent 在本地自主地存储、管理和发送一小部分资金，而不是把所有链上操作都依赖中心化服务或浏览器界面完成。

第二，是把 UniSat API 进一步升级为 UniSat AI Kit。它的目标是让 agent 能够更方便地访问 Bitcoin 及其相关协议和资产，查询链上数据和交易状态，并在必要时自主完成链上交易操作。

我认为这些都是更高级 AI 应用的基础设施。只有当 agent 能够安全地理解链上状态、管理资产、构造交易并完成执行之后，更多真正 Bitcoin-native 的 AI use cases 才会自然出现。

至于更具体、更前沿的 AI 实验方向，我现在不想过早给出一个狭窄的定义。AI 时代最大的特点之一，就是选择空间突然变得前所未有地大。很多真正重要的方向，可能并不会从今天看起来最合理的路线中出现。

所以我更希望 Fractal 成为一个开放的实验场，而不是一开始就把大家的想象力限制在几个固定方向里。

从 Fractal 发展的角度，我们会尽力支持那些真正有创造力、有工程能力，并且愿意探索 Bitcoin-native applications、on-chain intelligence 和 agent-driven on-chain operations 的团队。无论是资金、基础设施，还是生态协作，只要是足够优秀的想法，我们都应该想办法帮助它变成现实。

inAI 是一个非常好的开始。更让我兴奋的是，它可能只是 Fractal 上这一类实验的第一个信号。接下来真正值得期待的，是更多团队开始把 AI、Bitcoin、inscriptions、native assets 和 agent operations 这些方向组合起来，做出今天我们还无法完全预判的新东西。

E-2605 (Lorenzo’s Library)
Creative Commons BY-NC-ND 4.0