Python is no longer just a wrapper. For years, Python has been the go-to language for AI but always with a catch: it relied on wrappers and abstractions to talk to the hardware. That’s changing. New JIT compilers are giving Python native access to GPU-level performance, and it’s rewriting how we think about speed, control, and optimization in AI workflows.Have you tried any hardware-native Python compilers yet? What’s the biggest performance gain you’ve seen or the biggest bottleneck?