achimala/dream-loop ? reverse-engineered prompt

Build me a visually impressive 3D demo with a closed loop workflow, where the AI first dreams up a strong target screenshot, then builds toward it, then compares the live result against the target and keeps refining until it looks great.

I want something that feels polished and dramatic, with an artful scene, good lighting, strong composition, and little details that make it feel alive. Use Blender if it helps with custom modeling, and use image generation plus vision feedback to steer the look. If you can, have a separate critic agent judge the current screenshot against the target and give feedback for another pass. Keep iterating until the scene looks close and high quality.

Make it run well in the browser, and don’t ask me a bunch of questions, just make good artistic choices and get to a finished result. If you need to, look up current docs online while building it.