I spent some time earlier today laying out the stuff I need Seed to do. Either some of the recent things I thought of simplify the design, or I forgot something. Or maybe a lot of the fancy/complicated stuff I came up with is transparent to the user in terms of common actions, which, if so, that's good.
I think next I'll try to put together the data required for Seed to manage its own repo state. My plan is to develop the data and client code, write a first draft that appears compatible with it, and use the draft to manage the second version, and start using the second version on itself as soon as possible.
It's too hot to think any harder and I'm tired already. Good night.