The issue with the proposal is that many functions exist that don't throw but aren't marked noexcept, or which only throw under circumstances that you don't trigger. For example, the callee can throw on failed bounds checks, but the caller makes sure that indices are within bounds. Personally, I only use noexcept when a function has a wide contract, i.e. it succeeds no matter what, with no preconditions/assertions. noexcept(explicit) would be pretty much useless to me.
Also note that the usefulness of noexcept is dramatically overstated by some developers. You should probably pay attention to it on special member functions, swap(), and a few other key functions, but it has virtually no impact anywhere else.
>
I fed my idea in ChatGPT and asked it to produce one concise paragraph
> describing the benefits this new feature could bring, and it gave me
> back:
I'm not going to read that AI slop. Is this some out-of-season April Fool's joke?