Claude isn't getting dumber... · AI Automation Society

Claude isn't getting dumber...

So I've seen a few posts about people complaining about Claude. I'm not sure what's going on. But Claude is working crazy good for me, and the reason why might be because of your setup. So to help out, I'll share how I approach working with these AI models when coding.

Claude should always make the least amount of changes, no matter what. Don't let Claude make 10+ changes in one run.
Run tests on the changes that have been added, end-to-end. Ask Claude to test everything so it works.
Ask Claude, what could go wrong here? What are we missing after critical changes?
Run sub-agents for tasks and keep your main context window as small as possible.
Make sure your claude.md file isn't bloated. A bloated one will cause so many problems.

Try these things out and let me know how it goes!

18 comments

Claude isn't getting dumber...