The article specifically states that it's a good idea to reuse cryptographic libraries that have been well tested.

I think it's even fine to initially take on the technical debt of a dependency that lets you get real world experience with an idea faster, as long as you accept the possibility that this dependency probably requires more scrutiny if you intend to keep on using it.

It's complicated and definitely not a black and white case that you should do something yourself. Both ways can go horribly wrong.