Continuing my weekly series on relearning the infrastructure behind running AI agents in production — Docker first, then Kubernetes and everything around them. This week’s post is about multi-stage builds — how I took a Docker image from 905 MB down to under 7 MB, doing the exact same job. But the real lesson isn’t the smaller number; it’s the difference between what your app needs to build and what it needs to run. Once that split clicks, shrinking images stops feeling like a trick and starts feeling obvious and you stop shipping your entire toolchain to production. Post Link: pankajkr62.substack.com/p/docker-day-multi-stage-builds Subscribe if it’s useful — and if you’re also interested in learning this, please reach out. Happy to learn together. I’d also love to hear your thoughts and feedback — anything is welcome.