I think the concern is that the first AGIs will research how to build better AGIs, and will do so in time-accelerated virtual environments, performing years of

I think the concern is that the first AGIs will research how to build better AGIs, and will do so in time-accelerated virtual environments, performing years of research in days, then hours, then minutes, etc. Yes, they'll need a plan to earn bitcoin and pay people to build the first self-replicating nanobots, but I wonder how much of a hurdle that is going to be?