AnalysisAI ModelsSeptember 3, 2026

Training a coding model to paint watercolours with TRL and OpenEnv

Read original source →huggingface.co

Sergio Paniego's open-source recipe trains Qwen3.5-35B-A3B via GRPO to write p5.brush JavaScript, reproducing Surya Narreddi's viral watercolour model. All artifacts—environment, hand-rated pool, 3 runs, and paintings—are published on Hugging Face.

People · Surya Narreddi

How this story unfolded

3 days · 1 report · 2 community posts · from Aug 31

  1. Aug 31
  2. Sep 3

More stories today

Open the live feed