Shipping With Agents
Running an engineering organisation when machines write most of the code
When authorship stops being the constraint, specification and review become it, and an organisation that keeps its old review, ownership and measurement model just converts a coding bottleneck into a queue.
Merge rate goes up, cycle time does not move, and everyone blames the tooling. The tooling is fine; the operating model is not. This book redesigns the functions that authorship used to absorb: what a specification has to contain now, how to triage review by risk, what PR size and queue discipline actually buy, who owns code nobody typed, how tests become the contract with the agent, and which delivery metrics still mean anything once most committed code is machine-written.
What it makes operable
-
01
Redesign review as scarce capacity, with triage, size limits and a named owner per queue
-
02
Write specifications and tests that hold an agent to a contract
-
03
Measure delivery when most committed code was not typed by a person
-
04
Keep ownership and craft intact as adoption increases
Contents
14 of 14 published
Every chapter is free to read in the browser, cites its own sources, and stands on its own if you came for one decision rather than the whole argument.
- 01 The Bottleneck Moved More merged pull requests, the same cycle time, and where the hours went. 11 min
- 02 What a Specification Has to Contain Now Writing intent precisely enough that a machine can implement it. 11 min
- 03 The Standards Every Agent Inherits The repository-level context that decides code quality before a reviewer sees it. 11 min
- 04 Tests as the Contract With the Agent The suite as the specification an agent is actually held to. 11 min
- 05 Review as the Scarce Resource Treating reviewer attention as capacity to be planned, not goodwill to be spent. 11 min
- 06 Automated First-Pass Gates and Risk-Based Triage Moving baseline correctness off people, and routing low-risk change around them. 12 min
- 07 PR Size, Batching, and Queue Discipline Limits and batching rules that stop the queue from becoming the product. 11 min
- 08 Making a Revert Cheaper Than a Careful Read Rollback, flags and blast radius as the control that lets you triage review at all. 11 min
- 09 Ownership When Nobody Typed the Code Accountability, on-call and code ownership for generated change. 11 min
- 10 Security and Compliance Inside the Loop Putting controls where the agent works instead of downstream of it. 11 min
- 11 Measuring Delivery When DORA Stops Working Which metrics survive, which mislead, and what has to be added. 12 min
- 12 The Two Cost Lines Nobody Counts Seats and tokens are on the invoice; review hours and incident cost are not. 11 min
- 13 Juniors, Craft, and Not Deskilling the Team Building judgement in people who never had to write the easy version. 11 min
- 14 Adopting This Without Breaking the Team A sequence of changes engineers will not have to resist. 11 min
Evidence · 3 sources
Third-party sources behind the book's premise. Every figure in them belongs to the party that published it and is attributed to them in the text.
