ParentFull threadqarl·Might be that the vast majority of their training data is poor writing.View on HN