I have ported this into my real-world compiler however in the same re-implementation I'm also adding RVSDG and incremental compilation so it's gonna be a bit un

I have ported this into my real-world compiler however in the same re-implementation I'm also adding RVSDG and incremental compilation so it's gonna be a bit until it can be tested in practice. > Also: won't this type less programs than bidirectional type inference? I haven't come up with a situation where that'd be the case. I do suspect the opposite could be true if extended with things like associated types. But for now that's overly speculative.