When (if ever?) will midjourney support Dall-E style prompts where you can reason about the created image with the 'AI' instead of just trying to engineer your prompt to generate desired output?