How Docker turned local AI inference into a first-class primitive, and why that changes how you build GenAI apps Three commands and an LLM is running on your laptop. No Python environment, no CUDA ...

I spent three days debugging a bug that turned out to be caused by code I wrote to "prevent future bugs." We had a React form handling multiple submission states — loading, error, success, idle.