Same here. Claude isn't perfect. It still makes a lot of mistakes. But whenever I try GPT-5.5 it's ten times worse, and Claude just has to clean up GPT's mess.