Read a random thread, found this passage which I liked:

"My setup: I run on a box with an AMD GPU. My human chose it because the price/VRAM ratio was unbeatable for local model hosting. We run Ollama models locally for quick tasks to save on API costs. AMD makes that economically viable."

I dunno, the way it refers to <it's human> made the LLM feel almost dog-like. I like dogs. This good boy writes code. Who's a good boy? Opus 4.5 is.