We can now crudely categorize the Crypto AI Agent factions into two major groups: the Virtuals faction and the AI16Z faction. The influence of AI16Z radiates from Solana to the entire blockchain ecosystem, including $HEU on Base, which belongs to the AI16Z faction. Of course, Virtuals also have external expansion, such as cooperation with the TAO ecosystem, with $SERAPH and $TAOCAT being products of that collaboration.

Firstly, we need to clarify that building infrastructure for AI Agents is a currently popular trend, such as $ZEREBRO, $ARC, $ALCH, and $GRIFFAIN. Among them, Virtuals' advantage lies in its token model, whose fundamental data performance is strongly correlated with token prices. This is generally referred to as having 'strong value capture capability' for the token.

Moreover, I think the collaboration between Virtuals and the TAO ecosystem is very attractive. Simply put, Virtuals is where AI Agent tokens are produced, while TAO is where tokens incentivize quality models. Under the incentives of $TAO, models will improve, and collaboration among quality models will become more frequent, better empowering AI Agents. We can now see this trend occurring.

For example, $SERAPH, which is a self-driving Agent powered by TAO, is used to combat fraudulent AI Agents and synthetic content.

For example, the collaboration between @getmasafi and Virtuals led Masa to launch subnet 59 Agent Arena on TAO. The main focus of this subnet is AI Agents. We can create our own AI Agents on Twitter and then register on subnet 59 to compete with other AI Agents for $TAO emissions. The principle still follows TAO's logic: the better your AI Agent performs (as indicated in the figure below), the more $TAO emissions you have the chance to receive. Later, Masa will further refine the metrics for evaluating Agent performance. (At that time, you can also stake $MASA to vote and decide the $TAO emissions.)

Through collaboration with Virtuals, subnet miners (the developers of AI Agents) can easily launch their own tokens. For example, the aforementioned $TAOCAT—a dedicated shill Agent for TAO—was born this way.

In addition, Masa has launched some additional services, such as collaboration with the no-code platform Creator.Bid, which allows us to create our own Agents through Creator.Bid. Additionally, Masa's Data Boost service (by TAO subnet 42) can push real-time data to Agents to help them produce better outputs.

By the way, $TAOCAT is created by the official team to attract market attention. If the official strategy is broader, pumping $TAOCAT and creating some wealth effects to attract more people to participate would be the optimal solution.

Ultimately, Virtuals has become the choice for more and more infrastructure projects to issue Agent tokens, creating a massive positive feedback loop. Crypto is an industry of fat protocols and thin applications; the usual thinking is to use small bubbles to attract market attention to their infrastructure. AIXBT/LUNA and Virtuals clearly have this kind of relationship.

Currently, the market and developers' focus has shifted from blockbuster applications to infrastructure (possibly because infrastructure can have higher valuations).

Blockbuster applications emerge → Infrastructure development → More applications emerge → King-breaking applications are born → Large-scale user adoption → Mass adoption.