18 Weird and Wonderful ways I use Docker

Docker is being used as a lightweight “app sandbox” for everything from full GUI browsers to office suites, GPU-accelerated scientific workloads, and even a browser-based hacking lab—so tools don’t clash, and risky files get contained. The through-line is isolation: run each workload in its own container, map only the ports you need, and keep dependencies from stepping on one another the way they often do on a shared OS.

One of the most striking demos is running a full graphical web browser inside a container. By pulling a LinuxServer.io browser container and mapping host port 3000 to the container’s port 3001, the browser becomes reachable at localhost while still inheriting container isolation. Access is delivered via Chasm VNC, which provides a GUI session to the containerized browser. The same pattern then gets reused for Obsidian—Obsidian notes inside a container with GUI access—again relying on port remapping so multiple containerized apps can coexist without colliding on the same host ports.

The workflow expands into practical productivity and research tools. LibreOffice runs in a container with host ports mapped (3005/3006 in the demo), letting spreadsheets and documents open directly in the browser. Folding@home is also containerized, including optional NVIDIA GPU acceleration. That part requires enabling GPU support and installing the NVIDIA container toolkit so Docker can pass through the GPU; the demo includes troubleshooting steps like restarting WSL and verifying GPU visibility with nvidia-smi before confirming the Folding@home container is reachable on its mapped port.

For day-to-day management, Docker Desktop is presented as the visual layer over the CLI: it shows CPU and memory usage per container and can integrate with WSL2. Docker Desktop also supports extensions, including Poor Retainer, which adds an environment-management view inside the Desktop UI.

Security and safety get a dedicated segment with the Danger Zone. The tool takes potentially risky documents (PDFs, office files, images), converts them into a “safe” PDF, then reconverts to pixel data in a separate sandbox to strip out danger. It requires Docker Desktop to stay running, and the demo shows the conversion process while monitoring container activity.

The transcript then shifts from “useful containers” to “build your own.” A custom Dockerfile is used to containerize an AI tool called fabric, built from a Go-based Docker image, cloning dependencies from GitHub, installing packages, and setting environment variables. After building the image, the container runs fabric commands, and the user adds a shell alias so the same containerized workflow doesn’t require repeating long commands.

Finally, container security tooling and offensive testing show up. Docker Scout scans images for vulnerabilities and suggests fixes by updating package versions; the demo walks through resolving high- and medium-severity CVEs and notes that a dashboard view exists. For hacking labs, the demo uses Kali Linux in a container with an isolated Docker network, then deploys DVWA (Damn Vulnerable Web Application) into that network so the vulnerable service is reachable only through the Kali container. Docker Compose is used to declare the whole lab (network, Kali, DVWA) in a single yml file and tear it down cleanly with docker compose down.

The closing stretch experiments with other OS containers (Rocky Linux works; macOS-in-Docker attempts lead to trouble), spins up a Raspberry Pi environment inside a container (Raspbian buster light), and finishes with a containerized “IT tools” web app offering utilities like RSA key generation, QR code generation, safe-link decoding, and more.