Diffusion Models That Generate Neural Network Weights: Community POC 'nngen' Released
Specific-Tax-6700 · reddit · 2026-11-08
Redditor vagrillo proposes and prototypes using a diffusion model to generate neural network weights instead of pixels: since diffusion models map text to RGB matrices, swapping in weight matrices could let you "draw" a network from a prompt — e.g. one that guides a pen to draw a man, or plays Pong. Gen AI generating Gen AI.
Key points:
- Small POC built with help from z.ai's zcode and GLM model, open-sourced at github.com/vagrillo/nngen
- Training at 3000/4000 steps on a laptop CPU; preliminary screenshots suggest it partially works
- Author found no similar papers and asks the community for related work and contributions
- Includes a small webapp for testing the model
More from coding & agent
- A three-model subagent setup: one for communication, one for planning, one for grunt work — lxfater · 2026-11-08
- Open-source product-engineering repo ships /delegation and cross-harness subagent skills — PawelHuryn · 2026-11-08
- Subagent failures are delegation failures: brief with goals, not step-by-step instructions — PawelHuryn · 2026-11-08
- Hermes Chronos plugin adds time-travel checkpoints and rollback for Hermes Agent — Teknium · 2026-11-08
- A reusable agent prompt: build a live 4D quaternion Julia fractal renderer — arthurcolle · 2026-11-08
- Addy Osmani: Managing coding agents is just management — and it beats prompt engineering — addyosmani · 2026-11-08