So far, I have generated over 30 versions of the exact same screen with the exact same content trying to prompt it to generate the way I want. I have created guardrails, guidelines, detailed instructions, reference screenshots and everything I can think of. As of now, AI hasn’t gotten it right and I am losing daily credits.
At this quality of output, AI is at least 10 times slower than if I asked a UX dev to do it. At least -10X. The change I am asking shouldn’t take more than 15-20 min for a developer to do it. And I am struggling for the last 4 hours.
Does anyone face the same where they feel doing it manually would have been more effecient ?
Here are some of the (frustrated) prompt additions I am using now:
”…keeping the contents, design, element size and styling exactly the same as the screen attached. Repeat - output has to be a bottom sheet view, not a full screen view.”
”Do not change any of the contents, design, element size and styling, screen size, etc.”
Also, Stitch is currently unable to distinguish between a question and a request for screen design/modification/update. Even in the PRO mode, almost everytime I ask a question, it starts redesigning. At the tail of every question-only prompt, I need to add “Do not redesign any screens yet” - only then it doesn’t automatically start a redesign !
Hey @satwikb, thank you for sharing your experience. We sincerely apologize for the frustration and the loss of daily credits.
We are sharing your specific examples with our engineering team to investigate why these direct instructions are being ignored.