This is really cool.<p>On one hand, there are a few problems with the proposal that I doubt a human would do:<p>- There are a few errors like the first example, `type untaggedUnion = TypeA |: TypeB`, not really making sense (since the union is untagged, `TypeA` and `TypeB` don't really make sense)<p>- Repetitive, there's <i>a lot</i> of wordiness and specifics. In this case, since it's a formal proposal, this is actually a good thing. But I still think a human writing the proposal could achieve the same descriptiveness with less wordiness.<p>- "Related Work" there are several examples which are untagged variants, which is redundant especially because ReScript already has these. The only truly relevant ones IMO are TypeScript and polymorphic variants in OCaml, and then there should be a mention of OCaml or ReScript's existing untagged union implementation.<p>- The proposal cites one paper, "Publication: “Pattern Matching with First-Class Polymorphism”: This paper by Garrigue and Rémy (2013)..." which is completely made up<p>That being said, I think this is a genuinely great proposal, and the AI doesn't detract from that. I can definitely see AI being used to write many, many more of these types of proposals and formal documents in the future, and I think that's a good thing