In the world of Blockchain, the characteristics of openness and permissionlessness have become the source of its allure. The Bitcoin network allows anyone to access it freely, while the Ethereum platform enables developers to autonomously deploy smart contracts and create unique applications. This openness and composability is vividly referred to by industry insiders as "currency Lego."

However, when we turn our attention to the field of artificial intelligence, the situation is quite different. The current AI era still remains in a relatively primitive stage, akin to medieval alchemy. Major tech giants are guarding their massive "cauldrons"—that is, their large language models—in their secret laboratories. They invest vast amounts of data and consume enormous energy, hoping to concoct some sort of all-purpose "elixir."

This process is full of mystery, difficult to predict, and completely closed off. For the vast number of developers, all we can do is pay to purchase the API services offered by these companies, much like buying a few drops of their refined "elixir." However, we know nothing about the refining process and specific ingredients of these "potions." We cannot improve them, cannot mix "potions" from different sources, and cannot build our own AI ecosystem based on this.

In this mode, we seem to be trapped in the workshop of the "alchemist", reduced to humble apprentices. This is not only inefficient but also greatly suppresses the possibility of innovation. Looking back at history, the true scientific revolution originated from the birth of chemistry—when we were able to break down the material world into basic, understandable, and combinable "elements", we truly mastered the ability to create new substances.

Similarly, the AI field also needs to undergo a revolutionary transformation from "alchemy" to "elementalization." Only when we can deeply understand the internal structure of AI models, breaking them down into understandable and combinable basic units, can we truly unleash the innovative potential of AI and promote the rapid development of this field. This transformation will not only improve efficiency but also pave the way for the democratization and popularization of AI, allowing more people to participate in AI innovation.