@tokyo_0 @yaelwrites I think one of the most urgent problems with LLMs is how sycophantic they are. For some people, it really fucks them up. I mean, I think everyone in tech has seen someone go off the deep end after prolonged AI exposure. And, I mean, I think I'm immune, and I explicitly configure all models I interact with to say less and cut the sycophancy right out, but we don't actually know what long-term use does to folks. I'm intentionally using it in measured doses due to uncertainty.