As someone not in the know about cryptography, I know just enought that the words "custom cryptography" scare the hell out of me.
The most important thing to do with cryptography features is to stress test them ("spend $10 on verification for every $1 on implementation"), and that too has become a lot easier with frontier models.
I keep hearing about post-quantum being an issue though.
AI can roll you crypto far better than what humans have built by hand. They can literally test things to an extent that no human ever would.
"Don't roll your own crypto." sounds a lot like "Don't you worry about crypto. Let _me_ worry about crypto."
If you're that kind of person, then sure go for it! But if you're a top person in the field of cryptography you didn't need anyone to tell you if you were fit to make a cryptography library anyway.
And if you're not that kind of person you do need someone to tell you because the whole field is a minefield where subtle oversights break the whole thing. If you want to just implement cryptography as a fun learning experience that's great. But actual implementations that people will actually depend on need to be handled by experts.
Therefore many people need to check it, to ensure it is correct.
The crypto you rolled yourself is unlikely to have been extensively tested.
It's like alpha software. It may have no known bugs, but it hasn't been tested in the wild. Once it has been tested and the bugs fixed, because there will be bugs, you can release a stable version.
You shouldn't rely on alpha software for critical things, like keeping secrets, secret.
So yes, write a crypto algorithm, but don't represent it as working correctly if you don't know it is or force others to use it, and don't use crypto that hasn't been extensively tested.