In this video I show a workflow that honestly feels like cheating: I didn’t write a single line of the pull request. I opened a GitHub issue, added a simple “coder” tag, and an AI coding agent spun up its own isolated dev environment remotely to handle the entire fix. The agent reads the issue, implements the change, and opens a PR for me to review—without touching my local machine. The key idea is that Coder Workspaces isn’t the “agent.” It’s the infrastructure layer that lets you orchestrate agents at scale. I’ve got it self-hosted on a VPS, and once the tag is applied, Workspaces registers a task, comments back on the GitHub issue with a tracking link, and boots an environment for Claude Code to run. Because each session is isolated, you can run tons of issues in parallel (and even share the setup across a team) while keeping everything separated and secure. If you’re trying to move from “cool demo” to “repeatable agentic system,” this orchestration layer is the missing piece.