Probably by using fleets of diverse AI agents as startups to organize, research, simulate/prototype, refine, and propose novel ETI detection systems. This approach is already used in other domains, after all.
Whether it's a good angle, I don't know. But it's a perfectly reasonable one, methinks.
