Some of the tech and especially the platform they're building is impressive, but in terms of raw image generation quality from results I've seen and my own experience, I don't find it anything close to DALLE-2
If you zoom in on a DALL-E generated image the illusion falls appart and you can see it's a vortex of pixel that fuzzily creates a shape.
So far my experience with Stable Diffusion is that, no matter how basic the prompt, the ouput will be sharper and more coherent.