What are your experiences? I’ve tried using Auto, GPT-5 Thinking, Grok-4 Thinking, Gemini 2.5 Pro Thinking, and many others.
GPT-5 hallucinates too much for my liking, even without much context, only after a few prompts. Grok-4 sometimes just acts offended and refuses to respond, only does file changes and runs commands (but still without good results). Gemini just apologizes over and over without actually doing anything meaningful.
Now there is probably a big part of user-error involved, but I’ve created a very complex php app in the past months using GPT-5 Thinking model, without many issues.
Now I’ve tried creating a new web-app, even crafted the beginning prompt carefully using ChatGPT, layed out all the steps needed and asked it to start with small tasks. It sadly hangs up on the simplest things, at the very beginning of the project.
Could it be, because I’m using Remote-SSH plugin to work directly on the server? Or what could I be doing wrong? I’ve tried to incorporate as many tools and frameworks as the AI suggested, but also tried it without, as barebones as possible. It always starts to fail at a bug and goes into endless fixing loops, without actually fixing the issue.
Do you have tips on how to start a complex project from scratch and proceed going forward? Thanks a lot!
P.S.: I LOVE Cursor and the time we live in, could already accomplish so many functioning programs without any coding knowledge, but on the other hand it gets very frustrating trying to fix a simple bug for days without any results. Even starting from scratch, trying different approaches.