I have been banging on Cursor the last few days doing some quick Flutter dev work.
I have custom rules in place and I got warned about exceeding my limits while useing claude-4-sonnet so I switched to auto as it recommended.
The code it suggested and or created was marginal…
When I switched back to claude-4-sonnet it even corrected the code created while in auto mode
This seems to be a common occurence with me unfortunately. AUTO mode feels pretty terrible, I wish we could select properly which ‘auto mode’ models we use.
I mean I guess I don’t mind going over for quality interactions but sometimes it gets tiring because it generates decent code but then I have linting errors and it’s faster to just fix them myself .. why doesn’t it generate code without the errors …
I’ve literally never once had Auto successful edit something.