Calling what LLMs do just "reading" is, at best, naïve anthropomorphization.

Come on. It reads all the code in the world; then, you ask it to produce code; then, it gladly writes out something "new" which is a synthesis of what is out there, often with direct copying. It's more than reading.