>tell it to fix a bug>fix works>it broke something else>tell it to fix that>fixes it meanwhile breaking 2 more things>tell it to fix those>succeeds but the original issue is broken again>repeatthe naysayers are right it sucks
>>108664939Did you try telling it to not make any mistakes?
>you have reached your rate limit
>>108664939it is genuinely useful for low-sensitivity tasks that have obvious success/fail conditions and not much elsei think people dismiss it too quickly because their starting expectations are too high