Sure, I’ve been able to create an entire application by simply providing prompts that outline requirements, desired features, UX details, and so on. Initially, it doesn’t run properly after testing, so I feed it the error logs, and within a few hours, I can typically get the application working. However, over time, the quality deteriorates. The results become less reliable, and it sometimes invents requirements I never specified. This often forces me to open a new chat, reread the codebase, and start over from scratch.
I’ve also tried using the notebook and docs features, but they rarely produce clean, maintainable code that meets the requirements I’ve prompted. On several occasions, I’ve had to ask it to start fresh, which sometimes leads to decent outcomes.
Overall, I’ve found that providing clear ideas up front and seeking clarification before any code is generated yields better results.