Comment by sgtwompwomp

This is dope, is this kind of like Wafer.ai but for local models? As in a coding agent optimizes the kernels so the local model runs continuously better? Cause that is compelling if so. If it’s more simple that’s cool too