It seems to me that LLM's could massively improve reproducibility issues if journals would require that the papers be reproducible by model X using a standardized prompt in < N minutes, etc...