Terminal Contagion: How Microsoft's Command-Line Coding Agents Beat the Org Chart

Enterprise rollouts now move on terminal gossip. The strongest read on Microsoft's four-month Claude Code and Copilot CLI study is that the company did not deploy these tools — its engineers recruited each other, one visible keystroke at a time. Adoption tracked who saw whom typing at a prompt, not who reported to whom or which team had budget.

Call it terminal contagion. The mechanism is straightforward: a coding agent that lives in the CLI is, by design, a public act. A coworker can glance at your screen and see the workflow in three seconds, the same way they'd notice a new monitor. Compare that to a web IDE that opens in a tab nobody else can see. Visibility is the product.

This rewrites the rollout playbook. Procurement and IT should stop measuring success by license counts and start measuring peer-visibility metrics — how many terminals are observable, how many first-time users sit within line-of-sight of an existing user, how quickly a new cluster forms after one. The HN discussion surfaces the obvious falsifier: high-output engineers self-select into tools anyway, so a share of the lift is selection, not causation. Even granting that, the spread mechanism is the actionable signal. Get the visibility surface wrong, and the millions in annual token spend are underwriting a tool the org never actually socialized.