AI created changes, should it create PR's or work on my own fork

zzuegg: > In this example it forces me to split my instance data over multiple vec4 components where an additional mat4 would be enough. Ultimately, i think for no reason since mat4 is already supported and used by InstanceData you are repurposing a named vertexbuffer to carry an additional mat4? Maybe we should add support to custom vertex buffers instead