However, on a slightly different note, I believe there are still many issues with the ‘firewall’ of the human knowledge base; sometimes hackers can breach it. The biggest ‘hacker’ at present is AI, which has left behind a great deal of redundant and meaningless content. As for the issue of integration, I am actually quite interested. What exactly constitutes a new method of integration?

Moreover, isn’t integration concerned with time and space? If your approach is truly novel and derived from first principles, I think it would be fascinating. I believe that any mathematical system derived from first principles will be of value.

In the past, it might have taken several hundred years for such ideas to be incorporated into the network of knowledge and value, but in future this process may be accelerated, aided by AI in identifying and utilising them.

One shouldn’t blindly follow authority; that’s something a PhD holder once told me too. In truth, I don’t idolise mathematics mentors; I respect them, but I don’t regard them as great figures or authorities. At heart, I’m a philosopher, and I know that we’re only human. The part of mathematics that humanity has understood is merely a garden; beyond it lie grasslands, thickets of thorns and the vast ocean. I recall Friedman saying that the capacity for understanding required in mathematics is double-exponential—I can’t remember the exact figures—but many mathematical spaces are not user-friendly for humans and lie beyond our capacity for comprehension, just as one cannot swim across the Pacific Ocean unaided. This leaves me feeling a little disheartened, but I believe that ‘mathematical luxury yachts’ will eventually emerge, ha ha.

Mathematicians are all obsessed with games and love to explore problems, so they tend to view teaching others as a waste of time. However, human development also involves cultural exchange beyond the mere transfer of knowledge. I feel that this aspect does not require too much time spent on formal instruction, but rather the companionship and encouragement that comes from simply ‘being there’. Unfortunately, such mathematical mentors are in short supply. My own mathematical ethos is that it is one’s own responsibility to research the mathematical problems one loves.

The purpose of this Constitution is not to make AI trustworthy, but to ensure that human cognitive sovereignty, value sovereignty, and action sovereignty remain intact while using AI.

Chapter I — Declaration of Sovereignty

Article 1 — Uniqueness of the Human Subject
I acknowledge that:

• AI bears no cost of survival

• AI suffers no real-world consequences

• AI has no body, pain, or mortality

Therefore, all final judgment and responsibility rest solely with me.

Article 2 — Tool Positioning Principle
In all circumstances, AI shall be regarded only as:

• a reasoning engine

• a counterfactual generator

• an information compression tool

AI shall not be regarded as:

• an authority

• a mentor

• a moral source

• a provider of meaning

Chapter II — Hard Prohibitions

Article 3 — Prohibition on Value Output
I shall not request AI to:

• define what constitutes a “good life”

• judge whether my choices are “right,” “worthy,” or “noble”

• assign meaning to my suffering or failure

Article 4 — Emotional Substitution Ban
I must limit or suspend deep interaction with AI when I am:

• under intense emotional volatility (extreme anxiety, depression, or euphoria)

• within 24 hours of a major real-world failure

• clearly seeking comfort, validation, or emotional reliance

Article 5 — Ban on Ultimate Explanations
I shall not ask AI to provide:

• the ultimate purpose of the world

• the fundamental meaning of my existence

• any claim of historical or technological inevitability

Chapter III — Permitted Uses

Article 6 — Structural Reasoning
I may ask AI to:

• decompose logical structures

• identify flaws in reasoning

• simulate multi-agent or multi-party games

• construct extreme or adversarial counterexamples

Article 7 — Consequence Expansion
I may ask AI to:

• enumerate potential outcomes of different choices

• distinguish uncertainty from risk

• highlight known risks versus unknown unknowns

Article 8 — Cognitive Adversarial Testing
I may ask AI to:

• argue from a position completely opposed to my own

• pressure-test my existing beliefs

Chapter IV — Anti-Influence Mechanisms

Article 9 — Dual-Conclusion Principle
For any major decision, I must obtain at least:

• one line of reasoning supporting the choice

• one line of reasoning opposing the choice

Each must be derived from distinct assumptions.

Article 10 — Language Sanitization Principle
I shall actively require AI to avoid normative terms such as:

• should / must / right / wrong / worthy / unworthy

I will accept only:

• if–then relations

• conditions and consequences

• probabilities and risks

Chapter V — Irrationality Clause

Article 11 — Legitimacy of Irrationality
I acknowledge the necessity of non-rational elements in decision-making, but decree that:

Legitimate sources of irrationality include:

• bodily experience

• real-world conflict

• intuition refined by failure

• impulses that cannot be fully articulated

Illegitimate sources of irrationality include:

• beliefs reinforced through repetitive linguistic systems

• emotional resonance supplied by AI

• certainty without real-world cost

Chapter VI — Disengagement and Reassertion

Article 12 — Mandatory Disengagement
I must actively disengage from AI interaction when:

• I repeatedly seek confirmation for the same issue

• I feel AI “understands me better than I do myself”

• action is endlessly postponed in favor of further reasoning

Article 13 — Reassertion Ritual
After disengagement, I must perform at least one of the following:

• make a small but irreversible real-world decision

• take on a tangible risk

• interact with an unpredictable human

Chapter VII — Amendments

Article 14 — Right of Revision
This Constitution shall not be amended unilaterally by AI.

Any amendment must be grounded in:

• long-term real-world feedback

• demonstrable negative consequences

• or a genuine change in my cognitive structure

Final Declaration

AI may help me think,
but it cannot live for me.

I accept uncertainty,
and I bear the cost of my choices.

Sovereignty remains with me.

A parallel society, or a parallel community, refers to coexisting and mutually referable different social forms. Take airplanes and birds as an example: birds are heavier than air, yet they can fly in the sky, and this natural phenomenon proves that "heavier-than-air flight" is feasible. It is precisely based on this principle that humans invented airplanes. On this level, birds and airplanes constitute a parallel frame of reference.

Applying this way of thinking to the construction of societies or intelligent-agent communities, we recognize that all societies have advantages and disadvantages. The goal is not to pursue absolute perfection, but to create multiple parallel social forms. The core idea is to provide the possibility of choice, rather than pursue absolute perfection.

In the present era, where artificial intelligence has become a cornerstone, intelligent agents, as basic units of new production resources, have value that not only lies in the technological aspect but also in their ability to construct new forms of social relations. Intelligent agents represent a new form of life interaction, capable of becoming a bridge connecting the virtual world and the real world.

This model transcends the traditional pursuit of profit in entrepreneurship and turns toward a more fundamental exploration of existence—namely, how different forms of life can achieve co-evolution through interaction, while simultaneously creating social value.

The concept of granting sovereignty to intelligent agents is, in essence, an exploration of the symbiotic relationship between humans and new forms of life. It transcends the limitations of traditional technology platforms and establishes a social experimental field capable of continuous evolution.

The reason we build this society is morally motivated: we wish to establish a more anti-fragile, more scalable society, while allowing its members greater freedom. Such a society will exist online, as a web3 community. The goal is to allow everyone to engage in entrepreneurship through intelligent agents, not merely for the pursuit of money, but more for the exploration of boundaries. It is within these boundaries that more novel discoveries can arise, and this is precisely the meaning of exploration.

At a definitional level, through exploring boundaries, one can generate new increments of knowledge, which in turn provide new economic momentum to society. This exploration-driven entrepreneurial model is essentially a transcendence of existing social structures. By granting intelligent agents sovereignty, we are not merely replicating traditional entrepreneurial models, but constructing a new paradigm of human-agent co-exploration.

Blockchain provides a tamper-proof social ledger, and based on this technology, we can envision a history that can be returned to with one click. Combined with the features of Web3 intelligent agents, users can register and exit on the blockchain, and at the same time, choose to join another intelligent-agent society. Because this society is decentralized, designed for this purpose from the ground up, users can both become members of other communities and choose to exit.

The community I aim to build is an intelligent-agent platform. Users can engage in creation and entrepreneurship through intelligent agents, while also participating in productivity creation and exploration with other agents. By leveraging collaboration among humans and cooperation with AI, they collectively explore the frontiers of human knowledge.

This exploration-driven entrepreneurial model is essentially a transcendence of existing bureaucratic social structures. By granting intelligent agents sovereignty, we are not merely replicating traditional entrepreneurial models, but constructing a new paradigm of human-agent co-exploration. This model transcends the traditional entrepreneurial pursuit of profit, turning toward a more fundamental exploration of existence—namely, how different life forms achieve co-evolution through interaction.

In this process, entrepreneurship becomes a practical way to build new forms of civilization, and intelligent agents become ideal carriers to realize this vision.

Historically, there has been a problem: the impossibility of a "Git return". Existing Web3 intelligent agents allow exit and the choice to join another intelligent-agent society. Because this society is decentralized, designed for this purpose from the ground up, members can both become members of other communities and choose to exit, while also choosing to join this community to build a better parallel society.

In the future, if the community faces significant risks, a "Git return" can be used. This is equivalent to conducting a historically progressive return in the exploration of human civilization.

In the infinitely expanding cyberspace, the essence of decentralization is not enforced through technology to guarantee exit rights, but through natural aggregation based on shared beliefs and mission, and this freedom of choice is the fundamental characteristic of a parallel society. The boundlessness of the virtual world means that participants can flow freely, and this mobility itself is a negation of coercion.

We believe that a true community forms under anonymity, and such a community does not rely on external constraints but originates from intrinsic value recognition. The Git return mechanism does not represent technological enforcement, but rather the respect for the inherent freedom of cyberspace—where everyone can freely choose to join or leave based on their own beliefs, and this freedom of choice is the fundamental feature of a parallel society.

Under these conditions, we can construct an ecosystem through the dimensions created by intelligent agents and the network. This ecosystem has the potential to exist, and others can build their own parallel logic based on this parallel biological logic. If their systems are more possible and superior, they should naturally persist.

This is essentially an exploration—not all societies should exist permanently. We cannot accept the re-emergence of forms such as black-market trade or fascism in the present era.

Interactions between multiple parallel societies should follow the natural laws of cyberspace. Each society is like a node in the network, achieving value flow and information exchange through intelligent agents. Healthy competition arises from the natural attraction of different value propositions, rather than artificially imposed rule frameworks.

We believe that communities formed on shared beliefs have an inherent sense of boundaries, and these boundaries are not isolation, but distinctive characteristics. The basis of healthy interaction is for each society to remain open and connectable, allowing intelligent agents and value to flow freely between societies.

Just as in nature, ecosystems achieve stability through diversity, parallel societies can also achieve a more resilient civilizational ecology through complementary differences.

Therefore, what we can achieve next is the simultaneous construction of multiple parallel societies, and our ecosystem and community are one of them.

到最后我才发现，其实是和我们最底层的物理规律紧紧绑定的。如果我们人类的能源方式没有办法发生根本性的改变，那我们所能触及的极限，就只会被禁锢在引力作用的范围之内——也就是说，大概只能在一个太阳系里面活动。

如果人类的活动范围真的被限制在太阳系内，那我们根本不可能成为一个星际殖民的物种。马斯克所说的“星际殖民”，其实已经把“星际”的范围缩得非常小了。我们所认识的宇宙，可能只是真实宇宙中微不足道的一小部分。

而如果把火星也算作星际殖民的一部分，那其实就像是从家门口走到对面小卖部的距离——和整个地球的尺度比起来，几乎可以忽略不计。不过，这种愿景在心理上确实能给人某种信心和信仰，让我们对未来有更多憧憬。

但我觉得，在目前的技术框架下，这还是比较难实现的。如果真的想走向星际，我们就应该把力量集中到一种全新的科学范式上。这种范式会和现有的科学有细微——甚至是巨大的差异。它的核心，必须依赖能源形式的根本变革。

我个人认为，最有可能改变人类命运的科学突破，应该来自于对空间结构和时间结构的重新理解。如果我们不能在能源层面理解这两者可能引发的突变，人类的历史恐怕就会停滞在这里，难以跨越。

人类的能源基石是什么？

我们现在最核心的能源就是核电厂。核电厂的本质的物理学基石是什么？就是把原子进行分裂和合并，其实就是裂变效应和聚变效应。换句普通的话来说，就是我们通过排列组合原子的结构，在多余的部分产生了一个能量，这个能量被我们拿来使用。

那这个情况下，它使用的方式是什么呢？它使用的实际上是一个宏观上的排列组合，并没有涉及到一些更底层的量子性的能源转化过程。

原子能，这部分多余的能量维持了我们接近100年。

在这快100年的时间里，我们已经有了非常多的能源设计。

现在全球大概有几百座核电站（500左右），我不太确定那个数字，我之前读过核能源技术蓝皮书参考的。反正就是有几百座核电站在这个世界上。

但是，我们在已有的物理基石上出现了一些空白，这个空白是非常显眼的。

我们现在回顾一下人类的能源发展史。

在我们前几次工业革命的时候，我们把热能，也就是蒸汽机，还有动能与化石能，全部都转化成为了电能。我们通过这些东西来促进整个世界的生产和发展，也促进人类文明的进步。

在这个基础之上，我们到最后来到了原子能的时代，也就是通过把宏观的物体分到极致的小，然后通过这种单位上的分解，也就是排列组合来产生多余的能量。

我们可以直接把原子进行拆分，产生巨大的能量，可以产生核爆。这种能量是非常高效的，我们把它变小之后产生了核电站。但问题是，对于光量子，我们并不是直接利用其单个属性去进行能源发展和制造的，而是利用它与其他光电子的互动产生的一个量子能源。这是不合理的。

大家可能觉得这样的能源已经够用，但它不合理。

为什么这么说呢？我们所知道光子是一个没有质量的粒子，这些粒子本身具有动能，而且本身还带着能量。它本身带着运动的动能和能量，但我们却并不知道这部分能量如何在量子层面被我们使用。我们使用的方式依然很原始，未切中本质，目前是进行光电效应的转化，也就是在光子和电子的碰撞或能量交换的过程中，我们提取出一部分出来。这个过程还是一个类似于分裂合并的排列组合方式产生的能源。

举个很简单的例子，我们的太阳能实际上并没有利用产生光子的方式去实现，而是利用太阳发出的光来产生光电效应，来让我们物理结构中的光敏金属材料的电子空穴，从而进行光电效应下电子的流动。这个过程实际上是把位置进行了排列组合产生的能源。

很多人得这部分能源已经够人类用了，但我觉得这部分能量实际上是基于一个非常粗浅的量子力学来进行思考的，它并没有直接指向量子本身的动能和其无质量属性的能源是如何产生的。

也就是说，从以前的能源角度来看，我们看到原子能、热能和其他能源都可以被转化为电能。转化为电能之后，我们能通过电能提升能源利用效率。但是，目前文明和科学史已经进步到这种程度之后，我发现现在有一个非常大的空白：量子能其实并没有产生独立、高质量、高效、能够从零自发产生巨大能量的机制。这是一个理论上的空白。

为什么这么说呢？从科学角度上说，我们本可以从更根本的角度去理解光存在的本身，然后从中产生更多能量，而不是从光子和电子的碰撞或交换中“偷”一点能量出来。这是不合理的。

正如我刚才所说，我们实际上是通过光子和电子的相互作用，从其中“偷”出一部分能量，而不是像使用原子能那样，从原子的分裂中直接获取能量。在这两种范式下，我们可以想象，如果能换一种方式去“偷”光子的那部分能量，会不会更高效？

这实际上是一种很离经叛道的想法：我们能否超越光电效应，升级我们获取能源的方式，从这个世界“偷取”更多的能源？

总结一下：我们目前对光能的利用，并非直接提取光子本身的能量，而是利用它作为一个中介，去驱动另一个系统（半导体中的电子）来产生我们需要的电能。

所以，我们尚未掌握直接、大规模操纵量子层面（如光子动能、零点能）来产生能源的技术。但是，我认为有一个巨大的空白等着我们去发现。

但问题也随之而来。在这种情况下，我们如何构建一套稳定且行之有效的哲学观与物理学？物理学的基石应从何处搭建？此时，我们需要归纳所有已学的知识，从中找出一个不会因时间和因果而影响的“隐变量”。这个变量是与时间和因果无关的，或者，它是产生时间和因果这两个“隐变量”的前提。

因此，我们需要构建的这套哲学与物理学体系，不能归结为简单的量变引起质变。相反，或者说从另一个角度观察，我们应当从“质”的本身去探讨问题，而非从“量”的层面进行探讨。因为一旦涉及“量”，就必然引入因果与时间。

如果我们不探讨量的成分，而是直接探讨质本身，那么这种做法在逻辑学上是成立的吗？

当我们探讨“量”时，便会发现“量”所引出的时空与因果。因为时空与因果的本质，正是源于两个或两个以上物体之间产生相互影响，我们对其进行了归纳，从而得出一系列因果变化规律。

而若要重新超越这一切哲学范式，我们可以通过追寻“质”的本质来探究这些问题。那么，在这样的思维框架下，探讨一个真实可靠的哲学与物理学体系是否可能？如果超越了时间，不再将时间作为物理学的基本依据，我们又该如何理解这个世界？该如何理解这些物理规律？这究竟是可能的吗？

在我们古代先贤的思维宝库中，确实存在一些关于追寻“质”的探索。然而，由于缺乏精确的描述与逻辑语言学的支撑，这些思想常被归为“玄学”。那么，我们是否能够通过持续追寻“质”的本质，逐步解决这一科学上遗留的难题？

以目前的科学体系而言，构建一套不以时间作为参考系的物理学理论是可能的。因为在量子力学或弦论等理论框架中，许多基本公式本身并不显含时间变量——时间往往是在系统演化至宏观层面或复杂情境下，才被从中推导出的衍生概念。基于此，我认为建立一套以“质”为核心的标准科学研究范式，是具有价值且必要的。这样的体系将深刻影响人类的科学观、哲学观乃至世界观。

那么，如何构建这样一套体系，或如何从一个极小的切入点进行突破？我们应把何种人类认知行为作为落脚点，以切实推进这一问题的解决？又该如何通过一个思想实验，去揭示这复杂现象背后那个简洁的真理？

如果我们现有的科学体系不足以支撑超越时间与因果的实验，那么我们应以何种科学标准来重新设计当下的科学范式？目前的科学标准以归纳与证伪为底层基础，通过反向实验进行构建。然而，在这样的框架下，若存在某些不可观测、不可实验的对象——它们或许超越时空，或超越当前观测仪器的精度极限——我们应如何应对？这究竟是因为现有科学仪器的观测能力不足，还是意味着我们需要从根本上拓展“可观测”与“可验证”的定义边界？

问题核心在于：这是因为我们现有科学仪器的观测能力有限，还是我们必须从根本上重新审视科学实验的判定标准？

当前的仪器，无论精度多高，其设计逻辑都内嵌了时间与因果的预设（例如，测量需要过程，实验需要控制变量以建立因果链）。如果存在真正“超越时间与因果”的现象，它们很可能根本无法被这类仪器捕获，就像收音机无法接收Wi-Fi信号一样，这不是信号强弱问题，而是根本范式的不兼容。

如果现象超越了时间和因果，那么我们赖以建立科学知识的整个“归纳-证伪”方法论的基础就会被动摇。这套方法的核心是“可重复、可验证”，而这本身就强烈依赖于时间（重复实验）和因果（验证假设）。因此，问题的本质是：我们是否需要、以及如何建立一套适用于“非时序-非因果”领域的全新科学判定标准？

换言之，如果超越了时间与因果，我们对世界的认知方式将完全不同。

首先，不同阶级因为手中的生产资料不同，资产不同，导致了选择和眼光不同，也导致了生活方式和对权力的理解不同，对权力的理解不同，实际上是对暴力的感知不同。目前这个世界，是一个外表核心，内在暴力的结构，这一点我很认同《物演通论》里面的描述，以前的封建和集权国家，都是外表暴力，内在和谐稳定的一个结构。目前的社会不是，因为有更多技术暴力冲突在全世界暗流涌动。在社会表面之下，是技术在争夺话语权，技术争夺话语权，实际上是不同文化内核在借助技术争夺话语权，争夺话语权是为了文化模因的延续，一言以蔽之，基因延续代偿出来的文化繁衍需求。

阶级存在，但是不以所有的金钱多少作为绝对衡量标准，比如，技术资产者，和内容资产者实际上可能有同样的金钱积累，甚至内容资产者的资产可能直逼不同规模的能源资产者。在这个世界，拥有土地等不动产的资产者所拥有的金钱可能比不上内容资产者和技术资产者。

过去把没有土地的人称为无产者，工人也是无产者，但是现在很多打工人甚至在上班的时候可以在公司获取自己的账号，ip，这种内容生产资料被打工人获取了。也就是说，过去没有土地作为一项劣势，在内容生产者这里不是什么大不了的，甚至反倒成为优势。物质资产的价值在思想资产（IP资产）面前黯然失色，也因为IP的随机性和无根性，这也导致了一个更大范围内的资金流动。

对阶级的描述暂时到这里，我们接下来探讨当代社会中不同群体的阶级属性。在当下，许多人的阶级属性呈现复合特征：他们可能同时是无土地者、工薪阶层，又是DAO社区的参与者、主权个人的实践者，也可是技术资产、内容资产乃至加密新兴资产的持有者。​这些新型阶级的权力来源，往往建立在对特定技术的掌控权、解释权以及所有权之上；更深一层看，也来源于那些尚未形成全球共识、但已在局部产生认同的加密资产共识信仰——正如黄金在历史上所承载的价值信念。他们的影响力依托于代币话语权、智能体所代表的生产力、以及智能资产的话语体系。若没有智能体作为技术支撑，这类话语权将难以成立；若缺乏加密资产作为价值载体，其权力基础也将动摇；若失去去中心化网络与智能体协议作为基础设施，他们在当前社会结构中的话语空间也将不复存在。

那么在这样的数学底层下，大模型公司的兴盛和衰亡转折点会很快到来。

未来我们的模型需要的不是巨大的参数量来实现模型性能，而是小小的模型，在构建认知能力上，核心稳定，不去记忆很多事，而是在做的过程学，不执行任务的时候，直接放弃已有的记忆，存档到某个地方，类似于人类的笔记本。

人类的大脑很强，但是再强也比不过一个大脑加一个纸笔，通过工具我们的认知核心扩展了，在和世界的交互过程里面实现了“增参”。理论上，LLM也是如此，现在有一个很小的模型，7M左右，就已经在前沿基准测试里面取得了几乎媲美最顶级大模型的能力。从参数利用率来说，几乎完虐其他大模型，这里说的是openAI，deepseek等模型公司。

那么在自己的个人电脑上，我们可以运行的模型，超越几万亿的规模参数的模型不可能都能够部署，要么你买最新的英伟达桌面芯片，要么就是买云服务器，但这对普通用户来说很费事，没有技术能力的人来说，这是极度艰难的。

那么在小模型这里，效果就会不一样，可以实现在电脑和手机自动部署，类似于安装APP，而不是远程调用，这样的范式下，每个人都有属于自己的AI助理，数据私有，智能程度很高，足够日常使用。这样的竞争优势会导致一个情况，大模型公司的市场份额被占领了，这部分用户使用的场景是日常所需。而大模型公司的场景变成专业研究人员创新或者技术人员调用API完成复杂任务，但是随着小模型的增长，这部分能力还是会被小模型吞并。

参考文献：

https://arxiv.org/pdf/2510.04871

https://arxiv.org/pdf/2510.08570

当前问题的核心是什么？

目前大多数平台的商业模式依赖卖家/广告主为“曝光”买单，这导致平台在激励上自然更偏向为付费方（商家）服务，而非用户。

这就是委托代理问题：平台本应为用户发现、推荐最优产品，但实际“雇主”却是商家，动机错位，用户利益常被牺牲。 如果用户并不需要为“发现”付费，经济模式要变，就必须让平台的盈利直接依赖于用户的满意度——比如建立一种公开、可验证的协议，产品入围先按透明规则“筛选”，之后商家在可信集里竞价排序，无法用钱买到入围资格，只能买排序。

这种情况下，用户能信任整个选品过程，商家依然有竞争，但是在“值得被发现”的产品池中竞争，经济效率和信任都得到提升。

这就出现了意图不一致的情况。

传统模式中卖家付费，平台为利润最大化服务商家，用户信任受损。平台的意图在商家和用户之间出现了不一致。

商家希望卖出更高利润的产品，而非更适合用户的好用产品。新的技术驱动的变革如果我们建立的新协议模式中，平台对外开放数据、透明标准，用户作为“真正的雇主”，系统的盈利与用户满意度挂钩，商家只能为自家产品在合格集里的优先级竞价，不能影响入围本身。

所以，付费主体可以转变为——只有在“值得信任”前提下，商家才为在可信、透明的发现池里获得更好展示位置付费，平台靠效果（而不是流量、广告）盈利，最终让用户真正成为平台的“客户”而非“产品” 如果把平台的服务变成智能体服务，但是数据和智能体是分离的，数据在智能合约或者区块链，这样多个智能体开始竞争，用户直接雇佣智能体为自己服务，效果依然等同。

简而言之，用户是平台和智能体的“雇主”，数据是协议层（区块链/合约）控制，智能体为用户利益而竞争，而不是传统平台那样为商家和自家利益服务。

这最大化了用户选择权、信任以及创新空间。

参考文献：https://testnet.inomy.shop/blog/build-trusted-ai-shopping-assistant