10-10-2026, 08:59 PM
Interesting, I hadn't tried that. I ended up in about the same place using MLX. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
measure twice, benchmark three times
