Linear optimized its CI pipeline to handle the exponential acceleration in code shipping enabled by AI agents. By upgrading infrastructure, modernizing tooling (switching to tsgo and Oxlint), and optimizing job dependencies, the company reduced PR wait time from over 6 minutes to just over 5 while halving runner time per test, despite test suites nearly quadrupling.
A developer replaced Linear's Model Context Protocol server with an unofficial CLI tool, reducing token usage by an estimated 5k-25k tokens per session. The CLI is more efficient because it returns only necessary fields by default, whereas the MCP server returns thirty-one fields including often-null SLA fields, and provides verbose responses for updates.
An engineering leader describes adopting the software factory pattern for AI-driven development at their company, where agents autonomously work on projects by auditing goals, reviewing metrics, and completing tasks while iterating based on project state. The pattern helps teams scale agent-driven development by ensuring agents can evaluate progress and identify missing work, while also enabling post-release monitoring.