luckily we’re not doing any parsing ourselves, parsing and typechecking error messages in general i think can be exactly as good or better than elm has, and a better parsing algorithm can actually help to make this a lot less effort