Articles
Notes from shipping software with AI
For engineering leaders who want AI-assisted delivery they can check. Written from our own work, not from the sidelines.
Start here
Review capacity
The review queue is the bottleneck AI coding created
Why more code did not mean more shipped, and where the queue actually forms.
Instruction drift
Agent instruction drift in enterprise AI teams
How every developer ends up with a different AI, and what a shared rule set changes.
What was tested
AI coding checkpoint contract: make checks reproducible
What “tested” should mean before a pull request is allowed to say it.
Latest
-
Test From the Specification Before Generated Code Becomes the Answer
When generated code arrives first, later tests can inherit its assumptions. Start from observable behaviour and protect the checker’s independence.
-
Resumable AI workflows: what state must survive a human decision pause?
A resumable AI workflow must preserve more than the last message. Keep the workflow and stage identifiers, specification version, changed files, executed…
-
A Green Test Run Cannot Prove the Product Promise
A green suite can faithfully prove the wrong interpretation. The missing link is an independent, reviewable connection between product promises and evidence.
-
Loop Engineering 1: faster iteration does not replace a strong core
A Loop Engineering foundation is the set of stable inputs and controls the loop amplifies: intent, architecture, acceptance criteria, tests, state, decision…
-
Loop Engineering: The Outer Loop Is Where Accountability Lives
Agents can run the inner cycle of investigation, implementation, and checking. Teams must still own the outer loop where evidence becomes a…
-
A copy change and a payment migration should not use the same AI workflow
Risk based AI development routes work by blast radius, reversibility, sensitivity, and uncertainty. Low-risk edits use a fast lane.