Comment by agumonkey

There's videos about Diffusion LLMs too, apparently getting rid of the linear token generation. But I'm no ML engineer.