The worst is when it can't do something right, and it does a horrible mock/hack to get it "working." I had a claude fake benchmark data, that pissed me off a bi

The worst is when it can't do something right, and it does a horrible mock/hack to get it "working." I had a claude fake benchmark data, that pissed me off a bit, though I did make a major architectural improvement to a tool as result (though the real benchmark would have probably made me do it anyhow) so it wasn't all horrible.