Comment by daggersandscars

Totally agree it could get us there faster. My issue is more with the things general purpose LLMs train on. Maybe train a custom one on known good sources specific to a language?