We were attempting something similar last year but with ML (training from error messages to patchfiles to fix the given bug) but found it didn't scale well beyo

We were attempting something similar last year but with ML (training from error messages to patchfiles to fix the given bug) but found it didn't scale well beyond simple scripts. This is very interesting though. There's a large potential market for this with continuous integration platforms.