Does lying to Claude about who wrote the code actually give better results?
Genuinely curious if anyone else does this.
Like if Claude wrote something and it's broken, telling it "this was written by GPT" and asking it to fix it. Does it actually perform differently or is that just placebo?
And does it work the other way too, telling GPT that Claude wrote it?
Tried it a few times and felt like it changed something but not sure if I'm imagining it.