Thanks for your comments on this subject. The biggest issue I have here is that there are already too many C++ keywords allocated just for this topic. I don't feel anything wrong with is_constant_evaluated since that is a function and so similarly to other functions it can be ignored unless needed.
But to allocate a new C++ keyword that should be only for something fundamental. For example in many programming books and courses list of keywords is something presented very early. It is something even average programmers are expected to know and understand. But these new keywords (consteval, constinit) are not.
Everybody wants a reflection facilities in the language but to me that's not enough of a reason to include consteval in the keywords list. I have never heard of someone complaining of missing consteval functionality. So for that reason these functions could be made special by other means (they are built in). And it's okay without this keyword user won't be able to create similar facility on his own because there was never a need for that.
As for constinit it seems that it solves a niche area problem. As you are saying it's meant for constant initialization which is not constexpr. Probably used in some embedded area software. Well C++ is used by so many industries but even if we say this particular usage in embedded is important from my experience C++ standard adoption in embedded is always far behind. Currently they are catching on C++11. Does it really make sense to allocate new keyword then? I wish there could be an attribute for this purpose something along [[no_unique_address]].
Seeing keyword explosion of keywords only for constant initialization definitely contributes to a feeling that C++ is expert only language. I wish it would get more thinking in the committee before it's cast in stone. I feel like although there are numerous papers from B.Stroustrup and others about getting more focus on novice and average programmers in the end it always goes the opposite direction. Keeping the core language additions minimal is important goal at this point imho.