🔧 Why Self-Hosted Claude Code Was 15 Slower Than It Should Be

Update (2026-05-14). The SimpleEngine prefix-cache patch described in
Finding #2 is now upstream as
vllm-mlx PR #523, merged.
If you're on a recent vllm-mlx build, the fix is already there — no local
patching required. The walk-through below is still useful for understanding
what the patch does and why it was needed.
{: .prompt-info }

Update...