model selection by task has been surprisingly accurate for me. When I’m writing marketing content, it usually starts me with gpt 4o, but for brainstorming or outlining I often get claude. It saves me a step and the choices have been consistently relevant for what I need.
