Look at this conversation [https://chat.openai.com/share/61167da0-4190-4ac5-8b63-f25a4462ea12]
I've never had 4o (or any other open ai model) compile a macro that successfully imports into KM. I'm not sure if it's something I'm doing that wrong but it still seems wonky.
Plus, it often provides instruction for writing macros that either use hallucinated action options or outdated UI. There's a Keyboard Maestro GPT on the store that claims to be optimised for writing macros but it doesn't claim to be loaded with any documentation, so I haven't really thought to try it yet.
1 Like
I'd be surprised.
XML is an extremely strict language, and at least in my experience, current LLMs get it sort of kind of right, which is not good enough.