AI × Web3: Who is building the infrastructure for this era?

When a true shift in technological paradigms occurs, we often see a frenzy first, rather than a system. The wave of AI that we are experiencing is no different.

As a primary investor, I have always believed that betting on the deep transformative forces of the industry is far more valuable than chasing superficial narratives.

In the past year, I have encountered a large number of projects related to RWA, consumer finance, information finance, and more. They are undoubtedly exploring the intersection of the real world and on-chain systems. However, an increasingly obvious trend is: regardless of the path the project takes, it will ultimately need to integrate AI's collaborative logic to enhance competitiveness and efficiency.

For example, RWA, considering how to use AI for risk control optimization, off-chain data verification, and dynamic pricing is the future direction; consumers or DeFi projects that urgently need an excellent user experience also require AI to achieve user behavior prediction, strategy generation, incentive distribution, and so on. Other sectors also have similar demands.

Therefore, whether it is asset digitization or experience optimization, these seemingly independent narratives will eventually converge into the same technological logic: if the infrastructure does not possess the ability to integrate and support AI, it will not be able to sustain the complex collaboration of the next generation of applications.

In my opinion, the future of AI is not just about becoming "stronger and more widely used"; the real paradigm shift lies in the reconstruction of collaborative logic.

Just like the early transformation of the internet, it wasn't because we invented DNS or browsers, but because it allowed everyone to participate in content creation and turn ideas into products for the first time, thus giving rise to an entire open ecosystem.

AI is also taking this path: Agents will become intelligent co-creation bodies for everyone, helping you turn expertise, creativity, and tasks into automated productivity tools, and even achieve monetization.

This is a question that is difficult to answer in the current Web2 world, and it is also some underlying logic of my focus on the AI+Web3 track: making AI collaborative, transferable, and profit-sharing is the system that is truly worth building.

What I want to talk about today is the only project so far that attempts to systematically build the underlying operation of AI from a chain-level structure: Sahara.

The essence of investment is the worldview, recognizing the value system of choices.

My investment logic is not just the narrative of public chains combined with AI, and then seeing which team's background is better before placing a bet.

Investment, in essence, is a choice of worldview, and I have always been asking a core question: Can the future of AI be jointly owned by more people?

Can it leverage blockchain to reconstruct the value attribution and distribution logic of AI, allowing ordinary users, developers, and other roles to participate, contribute, and continuously benefit? It's simple: only with the emergence of this logic do I believe that such projects have the potential to become disruptors, rather than just "abandoned public chain +1".

In order to find the answer, I basically scanned through all the AI projects I could access until I encountered Sahara. The answer given to me by Sahara's co-founder Tyler was: to build an open, participatory ecosystem that everyone can own and benefit from.

This sentence is simple, but it precisely hits the soft spot of traditional public chains: they often serve developers in a one-sided manner, and the design of token economics is mostly limited to Gas Fees or governance, rarely able to truly support a positive cycle of the ecosystem, and even more difficult to sustain the development of an emerging track.

I am well aware that this road is full of challenges, but precisely because of this, it is an irresistible revolution.

As I emphasized in my previous article discussing the "Evolution from Web2 to Web3": the real paradigm shift is not about creating a single product, but about building a supportive system.

And Sahara was one of the most anticipated cases in my assessment at that time.

From Investment to 8x Valuation Follow-up Heavy Investment

If I say that my initial investment in Sahara was because it was doing exactly what I believe to be the true leading mission of AI – building an AI economy and infrastructure system. Then let me explain that within just half a year, I was eager to invest again at an 8-fold pre-round valuation because I felt a rare strength within this team.

Two co-founders, one of whom is the youngest tenured professor at USC, specializing in AI. I believe that the value of a tenured professor in an American university born in the 90s is not only in the academic field but also in the fact that at this age, they still have dreams, energy, and the determination to realize those dreams. Having known Professor Ren for more than a year, I have witnessed what it means to work over ten hours a day, with stable emotions and humility, as a genius.

Tyler, the former investment director of a certain platform, was responsible for North American investments and incubators, and his understanding of web3 goes without saying. He is astonishingly disciplined: he only sleeps in multiples of 1.5 hours, maintains his fitness no matter how busy he is to keep in shape, and doesn’t touch a single piece of candy to keep his mind clear, working over 13 hours a day. I once joked that he was a robot, and he simply replied: "I'm lucky to have this level of busyness today." His source of dopamine comes from making progress on projects every day; dreaming is his passion, and he doesn’t need any other fuel.

I am very fortunate to have met them, which has changed me. I have finally started to sleep more regularly as much as possible, my emotions are gradually stabilizing, and I am exercising...

So when someone says that Sahara received capital's favor due to luck, I always unreservedly add, "The pursuit of capital is an inevitable result." I vividly remember how difficult it was to secure primary market financing in this round, yet Sahara was being chased by investors in the primary market.

What everyone remembers is that a certain investment company invested in Sahara. Sahara has opened up an investment era for a certain tech giant's entry into the Web3 AI field, and winning the company's AI award is a significant reason for the investment. In addition, some heavily invested AI funds, national banks, and so on are also guests of Sahara. What you can see is a group of institutions that are more inclined towards traditional technology and industrial resources starting to quietly bet on AI × Web3 because of Sahara.

Capital will only pay for certain directions and execution capabilities - this is the positive feedback on the depth of Sahara technology, team background, system design, and execution ability.

This is also why it can produce some real and solid structural indicators:

More than 3.2 million accounts have been activated on the test network, with over 200,000 data platform annotators (millions in the queue). The clients they serve include several leading companies, and they have already achieved revenue in the tens of millions of dollars.

On this infrastructure chain, at least from "who will do it" to "can it be done", Sahara has gone deeper and more steadily than 99% of the "AI Narrative projects".

The Ultimate Challenge of Public Chains: Ensuring Continuous Benefits for All Contributors and Driving Positive Economic Cycles

Returning to our initial judgment logic: In a system that combines AI and blockchain, is there really a mechanism that allows every contributor to be seen, recorded, and continuously rewarded?

Model training and data optimization rely heavily on a large amount of labeled data and interactive support; conversely, if there is a lack of user contributions, the project itself has to invest more funds to purchase data and outsource labeling, which not only increases costs but also weakens the value-driven community co-construction.

Sahara is one of the few Web3 AI projects that allows ordinary users to "participate in data construction from day one." Its data labeling task system operates every day, with a large number of community users actively participating in labeling and prompt creation. This not only helps to improve the system but also invests in the future with data.

Through the mechanism of Sahara, not only does it improve the quality of models, but it also allows more people to understand and participate in this decentralized AI ecosystem, linking data contributions with rewards to form a true virtuous cycle.

A typical example is the Myshell project on a certain blockchain, which quickly built a high-quality dataset covering multiple languages and accents by leveraging Sahara's decentralized data collection and human-machine collaborative labeling, significantly improving the training efficiency of its TTS and voice cloning models. This also propelled its open-source projects VoiceClone and MeloTTS to gain thousands of GitHub stars and over 2 million downloads.

At the same time, users participating in data labeling also received token rewards distributed by Myshell, forming a bidirectional incentive loop between developers and data contributors.

Sahara's "permissionless copyright" mechanism ensures the open circulation and reuse of AI assets while protecting the rights of all participants - this is the underlying logic driving the explosive growth of the entire ecosystem.

Why is it said that this is a scenario with long-term value support?

Imagine if you want to build an AI application, you naturally hope that your model is more accurate and closer to real users than others.

The key advantage of Sahara is that it connects you to a vast and active data network—hundreds of thousands, soon to be millions, of annotators. They can continuously provide you with customized, high-quality data services, allowing your model to iterate faster.

More importantly, this is by no means a one-time transaction. Through Sahara, you are connecting to a potential early user community; and these contributors are likely to become the real users of your product in the future.

This connection is not a one-time buyout; through Sahara's smart contract system and rights confirmation mechanism, it enables a long-term, traceable, and sustainable incentive system.

Regardless of how many times the data is called, contributors will receive a continuous share of profits, with earnings dynamically linked to usage behavior.

But this is not just a revenue model for data labeling and model training phases. Sahara builds an economic system that covers the entire lifecycle of AI models, with built-in profit-sharing mechanisms at every stage after the model goes live, including calls, combinations, and cross-chain reuse, allowing value to be captured over a longer period.

Model developers, optimizers, validators, and computing power contribution nodes can now continuously benefit at different stages, rather than relying solely on one-time transactions or buyouts.

This system brings a compound effect for model combination calls and cross-chain reuse. A well-trained model, like building blocks, can be repeatedly called and combined by different applications, with each call generating new revenue for the original contributor.

Because of this, I agree with Sahara's underlying belief: a truly healthy AI economic system cannot be just about the plunder of data and the acquisition of models; it cannot be just about allowing a few to reap all the benefits. It must be open, collaborative, and mutually beneficial—where everyone can participate, and every valuable contribution can be recorded and continuously rewarded in the future.

But the closer we get to the real structure, the more challenges there are.

While I am optimistic about Sahara, I will not cover up the challenges the project will face because of my investment position.

One of the major advantages of the Sahara architecture is that it is not limited to a specific chain or a single ecosystem.

Its system was designed from the beginning to be open, full-chain, and standardized: supporting deployment on any EVM-compatible chain, while also providing standard API interfaces that allow Web2 systems—whether it is an e-commerce backend, enterprise SaaS, or mobile app—to directly call Sahara's model services and complete on-chain settlements.

However, despite the extreme scarcity of this architectural design, it also carries a core risk: the value of the infrastructure lies not in "what it can do," but in "who is willing to do what based on it."

To become a trusted, adopted, and integrated AI protocol layer, the key for Sahara lies in how ecological participants assess its technological maturity, stability, and future predictability. Although the system itself has been built, whether it can truly attract a large number of projects to land based on its standards remains uncertain.

It is undeniable that Sahara has achieved key validation: providing services related to data for several leading enterprises and addressing some of the industry's most challenging data demand issues, becoming an early signal of the feasibility of this system.

But what needs to be seen is that these collaborations mainly come from the Web2 world; what truly determines the long-term development of Sahara is still the entire Web3.