If OpenAI does it too, well then I guess my principles will be challenged. Gemini CLI is unfortunately too eager to use Flash despite my setting it to use Pro, and it generates ass code that doesn't work, so that's out. I might have to invest in some GPUs then and run local. We'll cross that bridge when we get there though.