It's using gpt4-turbo currently, was using gpt4 with fine result. The instructions are rather simple, so looking at what lama3 is doing, I think it could easil

It's using gpt4-turbo currently, was using gpt4 with fine result. The instructions are rather simple, so looking at what lama3 is doing, I think it could easily do it, I just don't have the knowledge and since the problem was solved with gpt4 I didn't want to put time into it. I'm hoping it will come with the community, with time But fine tuning a model will take this cost down 100x. Soon enough (couple of years) it will be compiling in the avg local machine. And just to top it off, the way compiler is built, compiling will be invisible so no more compile time