sometimes I reach some coding points not knowing if the agent is smarter than me and trying the best way to solve the issue OR IF IT'S TOO FUCKING DUMB AND JUST CAN'T FUCKING ADMIT THAT IT CAN'T SOLVE THE FUCKING PROBLEM

mago hacker 🇧🇷🏴‍☠️ building /gnars and /nounspace

claude 4 feels really dumb for some reason. I spent like $150 on it in Cursor over the long weekend and ended up throwing most of what it wrote out.

Hacker founder using ai to coordinate onchain economies /tenfold, nounish channel client /comint

More: rubinovitz.com

I wonder if we're finally seeing some overfitting from chasing benchmarks too hard

I think Claude 4 is overconfident LOL

but yeah, it keeps trying weird ways of solving the problem and I end up stopping it to do it myself because I'm just tired of waiting for it to understand