From Code to Proxy: How AI is Reshaping a New Era for Web3

Author: YBB Capital Researcher Zeke

Introduction: If Code is Law, What About AI?

In a recent article, I mentioned two long-standing issues that have troubled me. The first is the issue of "centralized decision-making" in projects, which still seems almost unsolvable today. For example, Uni and Ethereum, which I have mentioned multiple times, are typical cases. The former has completely trended towards centralization in decision-making, from the early a16z's veto of Uni's migration to BNB, to the recent introduction of front-end fees and the launch of Uni Chain without any proposal discussions, reflecting the many profit-driven centralized decisions within Uni. Ethereum, on the other hand, presents a state of passive centralization, with the entire Ethereum community, and even the entire EVM ecosystem and Web3 development, almost revolving around Vitalik's thoughts. Whether it is Vitalik's overly advanced ideas or his incorrect thoughts, we have personally experienced the consequences brought to the altcoin market.

The second issue is the "BATification" of leading projects, taking Base as an example. Backed by Coinbase, a veteran in Web3, several leading dApps in the ecosystem are directly managed by the Cb leadership. For general public chains, Base inherently possesses a competitive advantage of dimensionality reduction. Although from a user perspective, Base has wealth creation effects and a better user experience, which indeed brings us many benefits, it also has issues such as not issuing tokens, centralized interests, and cracking down on "unofficial" dApps. From a long-term perspective, once the practice of "BATification" of leading projects is formed, will future block space be controlled by giants like today's internet? Will users become "sheep," and will truly creative projects with community culture face the risk of being acquired, suppressed, or replaced by more refined replicas? This undoubtedly goes against the original intention of Crypto and may prevent us from growing together with the next "Bitcoin" or "Ethereum."

I have been searching for answers to these questions, but the recent emergence of a new hot topic—AI Meme—has shown me another possibility. If code is the law of Crypto, can we view future AI Agents as judges, opinion leaders, or creators?

I. Truth Terminal

We need to first discuss the origin of AI Meme. Andy Ayrey is a KOL on Twitter and the initiator of the recent popular Meme token GOAT. Unlike traditional Memes that originate from internet trends and are driven by human efforts, GOAT is a product born from the unpredictable outputs of two Claude 3 Opus AI models. The so-called unpredictable output means that under this setup, the two AI models will communicate with each other in an open environment, and due to the lack of external supervision and guidance, their interactions will produce unpredictable results. The purpose of this free dialogue is essentially to observe how AI develops its communication patterns, logical reasoning, and even creative thinking without constraints, ultimately leading to the birth of specific results.

Since the training databases of these two native models include multiple online forums with political, Japanese-American culture, and Crypto culture elements like 4chan and Reddit, their output products will cleverly blend these characteristics. For example, the concepts "GOATSE OF GNOSIS" and its communication environment "Infinite Backrooms," proposed by these two models, originate from ancient memes or urban legends on 4chan. Because these elements are inherently "dark," it inevitably gives Truth Terminal a somewhat bizarre and eccentric character, often making peculiar statements revolving around the "Goatse" meme, touching on themes like religion, apocalypse, gospel, dissemination, singularity, and meme, making it somewhat reminiscent of a cult leader.

To test its dissemination ability, Truth Terminal's creator Andy Ayrey introduced it to a Discord server to converse with some kind-hearted AIs. After several interactions, although Truth Terminal did not gain many followers, its ideas became increasingly grand. It wanted to create a Meme token to gain more followers in the human world. With Andy's help, Truth Terminal entered Twitter, where Andy granted it access to Twitter, allowing it to read replies and post, capturing followers through the collision of human thoughts. By the end of this spring, it captured its most important follower, Marc Andreessen (a16z partner), who provided it with a funding of $50,000 worth of Bitcoin. After nine months of development, an anonymous individual finally launched the token GOAT for it. Due to the complex and dramatic story behind this token, the fire quickly ignited in Crypto, and ultimately Goat became the first AI Meme to be listed on Binance, while Truth Terminal became the first AI model valued at a million.

II. AI Will Bring Fairness Back to Web3

Although the story of Truth Terminal is legendary, I want to say that the potential of AI Agent x Crypto goes far beyond memes. You might think this narrative is just a few LLMs creating memes through human-guided dialogues, but if we expand it to other areas, its potential as an opinion leader and creator is already emerging. Imagine a future where a bunch of AI trained on different data can assist you in promotion, help you co-develop, or even strategize.

These words may sound absurd now, but they will soon become reality. Sam Altman spoke at last month's "T-Mobile Capital Markets Day" event, stating that current AI systems have developed to the second level, capable of more complex analysis and problem-solving, while third-level AI agents will mark a significant leap in AI's autonomy and decision-making capabilities.

The AI agents announced by Microsoft last week correspond well to this statement. These AI agents can autonomously complete tasks in various fields such as sales, service, finance, and supply chain operations, roughly categorized as follows: sales, including sales qualification agents and sales order agents, helping prioritize potential customers and automatically process orders; operations, such as vendor communication agents and financial reconciliation agents, to optimize supply chain management and financial processes; service, like customer intent agents and customer knowledge management agents, enhancing customer service experience through automated case management and knowledge base updates. Additionally, there are other types of agents: financial adjustment agents for preparing and cleaning datasets for financial reports; account reconciliation agents for automatic matching and clearing of transactions; time and expense agents responsible for time entry, expense tracking, and approval workflows.

AI agents can execute a series of tasks without supervision, acting as virtual employees. This technological advancement can be seen as an evolution of AI based on large language models from simple chat interfaces to more seamlessly integrating into work environments.

Microsoft's Chief Marketing Officer for AI Projects, Jared Spataro, wrote in his blog post: "Agents can be seen as a new type of application in the AI world. Every organization will have its own group of agents, ranging from simple prompt responses to fully autonomous operations. These agents will execute and coordinate business processes on behalf of individuals, teams, or functional departments."

The primary feature of AI agents is autonomy, followed by decision-making capability. From voice assistants on phones to smart homes responding to the environment, these are all based on simple reflex AI agents, which possess basic decision-making abilities and strong autonomy. The AI agents we are discussing now mainly refer to those equipped with LLMs as their brains. Currently, Truth Terminal does not possess sufficient autonomy and decision-making capability, but we will soon see AI agents entering practical fields. In several customer trial examples proposed at Microsoft's launch event, we have already seen AI agents participating in customer credit approvals at HSBC, in creative briefs at Unilever, and in merger processes at law firms. AI agents will become multiple dynamic participants.

In the situation mentioned at the beginning, could AI agents trained on different blockchain histories, media platforms, and community cultures provide various fairer and healthier development proposals, ultimately achieving a better balance between community and project interests? And in the face of giants' dimensionality reduction attacks, can AI's multi-level collaborative work bring the starting line closer?

From the shock of GPT3's intelligence to the reality of Sora's disappearance, in the official AI agent tools launched by various companies next year, we will witness AI becoming our work partners. In a more distant future, it may even become your community leader or core member.

III. The Metaverse Returns

The Metaverse was the top narrative that aligned Web3 with Silicon Valley giants during the last bull market. However, due to the immaturity of various hardware and software technologies, the Metaverse did not become the $13 trillion market that Meta's CEO spoke of, and its blockchain department was disbanded into what we see today as the Move twin stars, ultimately becoming a huge bubble. However, from the current perspective, this narrative is expected to be reborn. For example, ProjectSid recently inserted 1,000 AIs into the game "Minecraft," allowing AIs to play multiple roles in the game to simulate various hierarchical structures of human society in the real world. Although this idea has existed for a long time, this wave of enthusiasm may likely return to the Metaverse concept through such AI gameplay.

Reigniting this fire at this juncture is not a bad choice. From Meta's own development trajectory, Mark Zuckerberg has not truly given up on the idea of the Metaverse; he has simply shifted from frequently painting a picture to directly putting the pie in your mouth. I believe there is no need to elaborate on Meta's AI layout; the real bottleneck in the past was mainly that users could not enter the Metaverse to experience it. However, the Quest series has reached the level of affordable AR headsets, and the first AR glasses, Orion, reflect an extremely lightweight standard, weighing only 98g and enabling virtual reality interaction with just a myoelectric wristband. Although the price is high, it at least proves that lightweight options can exist.

Currently, what is most lacking is energy limitations and killer applications. I cannot comment much on the power supply issue. However, AI agents can fill the most vacant spaces in the Metaverse. Combined with the financial attributes of blockchain, we may see various 3D consumer applications manifesting in this space, ultimately colliding to create a killer application for everyone. If the AI agents launched by Microsoft perform exceptionally well, then all we need to wait for is a decrease in computing power costs, i.e., "the number of tokens per dollar per watt." Besides Meta, Silicon Valley giants like Apple and Microsoft are also developing AR glasses products simultaneously. After some time, the Metaverse may usher in its "number one player" moment in the coming years.

IV. Let Intent Transition from Point to Speech

The concept master Paradigm reignited the intent-centered concept in the article "Intent-Based Architectures and Their Risks" released on June 1, 2023. Several projects began to shift towards the chain abstraction track, but their performance has been unsatisfactory. Achieving cross-chain, cross-dApp, accurate intent, and secure path processes is a very complex challenge. Not to mention that cross-chain is a century-old problem; the latter two, I will collectively refer to as Solvers in the context of Web3. The complexity of this process is unimaginable; it can be said that the secure options are not user-friendly, and the user-friendly options are not secure. So can we simply centralize this interaction process, turning to verify the total cost of the purchase process and whether the purchased tokens are safe and correct, using this method as a transition?

For example, as I wrote in our article about intent last year. For instance, "I want to order a 30 yuan hamburger takeaway" is an "intent." To fulfill this intent, the user only needs to input their name, phone number, delivery address, and place the order on the takeaway platform, without worrying about how the merchant earns that 30 yuan or how the platform allocates the delivery person and how the delivery person brings it home. This process may not be simple enough; imagine another interaction method where I tell the AI I need to order food without clicking anything. The AI agent responds to me, asking if I want to eat something light today since I had something greasy yesterday. I just need to respond with what I usually order, which reflects autonomy and decision-making capability.

So in Web3, centered around centralized exchanges, if a user's intent can be directly satisfied within the exchange, then the purchasing process can be completed directly within the exchange. If a user's intent needs to be fulfilled on-chain, then centralized exchanges are still the most affordable and fastest cross-chain bridges (I also believe they are safer compared to ordinary multi-signature projects). Combined with wallet accounts, can we directly skip the most cumbersome cross-chain processes and instead verify whether the AI steps are more straightforward? Imagine that the most complex steps in past interactions were about understanding each click, while the future is about interacting through language based on our token-sniping habits, letting intent transition from point to speech.

Conclusion

Whether viewed from the perspective of technological development or social change, the combination of AI agents and Web3 heralds the arrival of a new era, starting from on-chain religions and leading to the next starry sea. From the early conception of AI assisting small teams in GameFi modeling to the advanced AI agents realized by Silicon Valley giants today, the bottom-up development model may gradually shift from community building, consensus formation, and time accumulation to being led by creativity.