Chapter Thirteen · Working in a Team
Working in a Team
The book's discipline, moved from one engineer to a team. Five topics review a colleague's AI-assisted pull request, put an automated reviewer on it and count what it caught, write down what the team had only assumed, answer who owns what comes back, and hand an issue to an agent that nobody watches.
Everything so far has been one engineer and one agent. A team adds people who did not see the session: a reviewer who reads a pull request the agent drafted and the author only partly read, a colleague whose plan has different data terms, a client who wants to know which code was generated. The habits do not change. Who applies them, and where, does.
This chapter follows Kai's next change through Cragmoor. Ines reviews it the way Chapter 6 taught you to read your own diffs, with the author in the room to answer questions. An automated reviewer reads the same change and the page counts what it caught and missed. The team then writes down the policy it had only assumed, faces the question that policy raises about who owns generated code, and ends on agents that take an issue and return a pull request on their own.
The thread through all five topics is accountability. An approval is a claim that someone understood the change, and a tool cannot make that claim for anyone.