The process is basically `rbs prototype` or `typeprof` to generate a sig file and steep then uses that. TBH it's not that much of a chore as it sounds and already gives you good-ish results. That gives you a bunch of material to work from and progressively refine types.
Speed is very good, I didn't notice anything slow.
In the Python libraries I maintain, types are useful to anyone who installs the default Python VS Code plugin, because it gives them intellisense (but doesn't give them red squiggly lines for type errors) even if they do nothing to set up types or a typechecker.
Is that or will that be the case for Ruby/Steep?
We plan to do that at some point for the Ruby Datadog tracer public API, so that users can have intellisense and don't need to jump to the docs.