Another article with a similar concept, around AI, but not LLMs: https://blog.jtoy.net/choosing-the-appropriate-abstraction-l...