i'm really struggling to get any effective results out of this AI coding agent at the moment. gonna try seeing if cursor does a better job.

the AI is just not doing anything write. it isn't using existing event caches in the test files, nor is it using the correct method to construct and sign events, and i can't understand what it's trying to do with the code it's writing, it's not right at all.

i probably would have fixed at least one of the bugs i've got on my agenda by now if i'd actually just sat and wrote it by hand.

Reply to this note

Please Login to reply.

Discussion

Reminds me of people knowing their codebase vs newbies: https://threadreaderapp.com/thread/1943360399220388093.html

I think that instead of developing ever better AI, we need to develop something akin to a martial arts of the mind and give ourselves disciplined super powers.

For instance if I could hold an arbitrary set of data in a 1000 slot mental stack I'd be unstoppable. Heck even 5 level stack would be amazing. Mental math breaks down for me when I have to push more than 2 numbers into the stack.

Cursor is the best AI tool for me, but you need to be very very very specific about what you want. Btw im only using Claude models