In hindsight, I think you are right, let's introduce a new type called
"runtime_tuple". I will update my proposal as well.

On Fri, 17 Apr 2026, 9:11 pm Jason McKesson via Std-Proposals, <
std-proposals_at_[hidden]> wrote:

> One more thing: doing a tag for `std::tuple` is an implementation
> non-starter. Right now, code exists that looks something like this:
>
> ```
> template<typename Ts...>
> void func(std::tuple<Ts...> &tpl)
> {
> //Iterate over Ts...
> }
> ```
>
> This code relies on `Ts` being the sequence of types actually stored
> in `std::tuple<Ts...>`. And this code in particular relies on template
> argument deduction to allow users to just pass in a `std::tuple`.
>
> Making one of the `Ts` into a tag type breaks this code. And every
> piece of code like it (such as `std::apply`).
>
> Can it be fixed? Sure; it can check for the tag. But breaking
> everyone's code is unpleasant.
>
> Furthermore, if `Ts` are all trivially copyable, then you can assume
> that `std::tuple<Ts>` are also trivially copyable. And thus do
> trivially copyable stuff with them. But your "optimized" type doesn't
> allow for trivial copyability regardless of the `Ts`. So it breaks
> there too.
>
> Just make a new type.
> --
> Std-Proposals mailing list
> Std-Proposals_at_[hidden]
> https://lists.isocpp.org/mailman/listinfo.cgi/std-proposals
>