Is there anything promising out there?
Is crowd sourced training still unfeasible?
I remember how fast the diffusion world moved in the first year but it seems it's stalled somewhat compared to first midjourney then Dall-e 3. Is it the same with text models?