> Subways are pretty efficient systems but that doesn't mean they are anywhere close to the most efficient system, especially considering they were invented during a time where it would be unthinkable to dynamically compute optimized schedules.

Personal rapid transit systems, which is basically what Loop is and sounds like what you're proposing, have been proposed several times and even built in a few places (Morgantown's is perhaps the most notable, at least in the US). They just haven't shown themselves to be very capable.

Subways are probably the most efficient system possible. They're a 1-D system, with transfers between separate 1-D system. When there's no conflicts between different lines (i.e., everything is grade-separated), you can dedicate most of the space to the actual trains themselves and the safety buffers for moving trains to stop. As you move to more complex systems, now you need to have extra gaps that other trains on other lines can merge into, which reduces scheduling flexibility. In other words, subways are much simpler, and it turns that simple systems are often the most efficient in practice. (If you want to see an example of what can go wrong even with computerized scheduling of dynamic routing, look at Denver's baggage system).