Is ChatGPT Getting Worse for Programming Help?

I switched to Gemini a couple of months ago and found that ChatGPT isn’t as reliable for coding tasks anymore. It feels like it’s mostly good for drawing now rather than providing useful programming advice.

Regarding how to program it better: It's all about figuring out how to prompt it effectively. The model doesn't seem to be declining; it's just a matter of refining the prompts or ideas you're putting forward.

Honestly, I’ve had varying experiences. Sometimes it seems to skip several steps on simpler tasks, but then it’ll surprise me by nailing more complex tasks just right. It can be unpredictable, but I wouldn’t say it's consistently getting worse. By the way, are you following test-driven development (TDD)? That might help.

It seems like OpenAI is tweaking how their models function constantly. This change can affect response quality and detail. When I hit a wall, I usually try altering my prompts or breaking them down into simpler parts, and that sometimes helps.

I've noticed similar issues in the last couple of weeks, too. While I find it still works better than Claude or Gemini for coding, I've gone back to using GPT-4 as it seems to handle C# coding much better than the mini or high versions. When GPT-4 gets stuck, I switch back to older models like 3 or 4-mini, but the high version doesn’t have the same quality it used to.