Is There a Recursive Form of Meta Learning?

I like the way you describe the evolution. I remember at some point I gave a talk about meta learning. Jürgen Schmitthuber was in the audience and he essentially called me out, like he likes calling people out. But specifically for your question on this more recursive form of meta learning, I don't think I agree that that has to be a requirement for these algorithms. And from that perspective, I can certainly imagine something that learns at more than two timescales that learns at different intermediate levels of timescales as well. but I'm also not sure if there's a lot of practical value from going even more meta.