← Back to context Comment by qarl 4 days ago Might be that the vast majority of their training data is poor writing. 0 comments qarl Reply No comments yet Contribute on Hacker News ↗
No comments yet
Contribute on Hacker News ↗