Introduction

The previous lesson left you with a feature implemented by an AI-driven workflow, but the work of checking it against the plan is still ahead of you. Running that review one pass at a time works, but it’s slow, and nothing stops you from wanting several angles of attack on the same codebase at once — as long as they don’t step on each other.

Conductor is a macOS app built for exactly that problem. Rather than switching branches or juggling terminal tabs, Conductor gives each AI task its own git worktree: an isolated working copy of the project checked out from the same repository, so multiple agents can read and write files simultaneously without conflicting with one another. Download it before the demo so you can follow along.

In this lesson, you’ll use Conductor to run several review tasks against the implementation from Lesson 1 in parallel, each one in its own worktree. Once a task’s work is ready, you’ll use Conductor to open a pull request straight from that worktree, review the diff, and merge it back into your project — all without leaving the app.

By the end of this lesson you’ll be able to:

• Explain how Conductor’s use of git worktrees lets multiple AI tasks run in parallel without conflicting.
• Set up and run multiple parallel tasks against the same implementation in Conductor.
• Use Conductor to create a pull request from a task’s worktree.
• Review and merge that pull request directly within Conductor.