I'have been dreaming about using a Diffusion Model that will generates Weights instead of pixels.
Diffusion models actually are vast used in order to generate from a natural language text description into images, and what is an image? a three dimensional matrix of value representig luminosity of red , green and blue. What if instead of matrix of colour intensity we train a diffusion model in order to create a neural network, and maybe a drawing NN too.
If it works you can ask something like:
"create me a neural network that will guide a pen in order to draw a man"
"create a NN of a game field where i have two paddle on left and right side of the field and a bounching ball..."....to create Pong

and in the future more complex scenarios!

I made, with the help of z.ai zcode and their fantastic GLM model a small POC that i published on my github:
https://github.com/vagrillo/nngen

It's a very small experimental implementation , but somenthing seems to work!

I have not find paper on similar implementation ad I'm asking the community to help me to find one, and to have contribution on this .

I also made a small webapp in order to test the trained model (actually i'm only on half training path 3000/4000 steps finished, but i'm using my poor laptop CPU at the moment)

This is some screenshot of preliminary results. (the best results achieved, in truth)

https://preview.redd.it/kje7wxpkac0i1.png?width=654&format=png&auto=webp&s=32c15d73aa4ea9e1460cce6b1947b850b7ee31bf

https://preview.redd.it/86uno8kmac0i1.png?width=659&format=png&auto=webp&s=5e20d5c3b481342c77746339d237dbb3ebd0738e

Italian explanation on my YouTube https://youtu.be/wAMO3Db-JoQ