There's some fantastic ideas at play here.

That idea of separate subnets for the true and false sides of the conditional node, which are allowed to produce differently typed outputs, is rather fun. I'm curious where else you're allowing subnets to be created.

If the live probe is based on selection, what happens when you select multiple things at once?

Also, I'm curious why the probe is a dedicated window, rather than every node showing live values all the time. Is it about visual noise / space? Or perf? Or some other tradeoff?

Thanks ivan, subnets are also used for loops (reduce, filter, map, for..) where the subnet is executed multiple times. For lambdas and shaders, where it's executed "at a later time" and similar to condition, we also have match (switch statement) and inspect, where only one subnet of a list of subnets is executed. The idea came naturally from regular programming languages, where you also have scopes.

Usually a ui can have a selection but still has also a single focus (where the keyboards at). I forgot to mention, that if you drag the probe window, it will stick to the node until you close it. Not sure if that's final yet, as sometimes you drag the window to get it out of the way.

Mostly noise and space. I want to keep the graph really tight. There will be a dedicated panel for more visualization.