Author: Daniel Barabander, investment partner of Variant Fund; Compiled by: 0xjs@金财经

What are intelligent agents good at? We discussed internally and concluded at least four points:

1. Close to human needs; 2. Help promote human actions; 3. Integrate and synthesize information; 4. Provide entertainment

First, be close to human needs. The agent is capable of processing human language, so in theory, any application that a human can use as a user, the agent can also use. But unlike human users, agents are able to provide services to other users on these platforms at scale.

As a result, agents can act as a layer on top of existing user-favorite applications, extending their usefulness. Take the Bounty Bot on Farcaster as an example. Users could have posted bounties externally, but that would have made the operation more difficult. It will take time to wait for the Merkle team to develop this feature. The bounty robot is a service provided on this basis.

By being close to where users are (meeting their existing usage scenarios), agents provide convenience, practicality, and open up ways to capture value in existing applications. But be warned: not all apps are equally good at supporting agents — the best apps to build on are those with a solid application programming interface (API), like Farcaster.

As a side note: I wrote a paper on one of the major legal concerns for agents on the Web2 platform - the Computer Fraud and Abuse Act (CFAA). In short: my research shows that if a user has full control over an agent's login credentials and the Web2 platform bans the agent, the user needs to stop running the agent. My thesis emphasizes building agents on a coherent/open platform like Farcaster, which is another reason why I am particularly interested in agents on this platform.

Secondly, help promote human action. Humans are great at signaling (indicating intent, etc.) but have trouble executing (especially in a networked environment). Agents bridge this gap by doing the heavy lifting, while humans guide the outcome through preferences.

A good example of an agent that takes advantage of this is BottoDAO. It creates works of art based on the opinions of Decentralized Autonomous Organization (DAO) token holders. Artificial intelligence takes care of the hard part of creating art, while humans’ preferences expressed through voting on artworks guide its creative direction.

Third, integrate and synthesize information. Agents are capable of processing massive amounts of data that are far beyond the capabilities of humans. For example, a trading bot analyzes a large number of on-chaindata to make decisions. Others, like @aixbt_agent on Twitter, mine valuable information from CryptoTwitter (CT).

Finally, provide entertainment. In the field of encryption, this may be the most rapidly developing category of agents, such as Truth Terminal.

Of course, much of the entertainment brought by agents on social platforms lies in the novelty of robot-generated content. But I’m more interested in bots creating entertaining content on their own merits, interacting with other users in interesting ways just like any influencer on the platform would.

The great thing about agents as Internet celebrities is that, like traditional Internet celebrities, once they attract a loyal audience, they can easily provide other agent services, especially those that are interesting to agents. A service that is easier to monetize directly than off-chain advertising.