💭 so deepleffen emulator is still dead because the company I finetuned on seems to be falling apart and the relevant models giving clearly broken outputs with or without finetune. But maybe I can use an ensemble of basemodels with parameter finetuning and a powerful manager llm who can hopefully 'get' the sense of humor to automatically pick out gems from the pile of crap.

I wish I could have a bunch of me try different things and then pick the one that worked out best at the end of the day - or just have a bunch of me.