Interesting post here about evolution of WIT and that maybe it should be primarily a machine readable format: https://blog.yoshuawuyts.com/rethinking-wit-as-a-machine-format/
I think this is pretty consistent with the approach we've taken with contract interfaces. They're primarily machine readable, but have human readable representations.
I think if this idea continues this is where something like WIT will differ from IDLs like protobufs, where the human readable component is more important.