Claude Code completely destroys ChatGPT for programming. I had this same kind of discussion with a bunch of my programmers at the beginning. They were using all kinds of AI products (other than Claude Code) and I basically had to force them all to switch. But they've all seen the results now.

Another point a lot of people don't get: Claude (the LLM) and Claude Code are different things. A bunch of my PeerVerity team programmers were using Claude with Github Copilot licenses and only paying $10/month, so they were also wondering why I was so forcefully pushing a $200/month license. But Claude with Claude Code performs at an entirely different level, because of the instrumentation around it provides.