I've been working on a small 2D graphics framework for the web and just put a first prototype online. It aims to make authoring graphics pipelines easier without hiding how the GPU actually works.

The core idea: you write your renderer as a function that returns a RenderGraph - a graph of resources (textures, buffers, etc) that describes how data flows through the GPU. The engine maps your graph to physical resources on the GPU in an optimised way.

The value proposition: For beginners, it could serve as a gentle onramp into GPU programming. For experienced developers, it could be a fast prototyping tool for experimentation.

I'm curious whether people see potential in a framework like this. I'd be very grateful to hear your thoughts!