Contracts feel like the right direction but the wrong execution timeline. The Ada/SPARK model shows how powerful contracts become when they feed into static verification — but that took decades of iteration on a language with far cleaner semantics. Bolting that onto C++ where UB is load-bearing infrastructure is a different beast entirely. The real risk isn't complexity for complexity's sake — it's that a "minimum viable" contracts spec gets locked in, and then the things that would actually make it useful for proof assistants become impossible to retrofit because they'd break the v1 semantics. Bjarne's concern about "incomplete" is more worrying to me than "bloated."