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
-
Comprehension Debt: When the Team Owns Code It Cannot Explain
A team can merge correct code and still become less capable of repairing it. Comprehension needs deliberate controls in AI-assisted delivery.
-
Loop Engineering 4: a human gate is not another permission prompt
A Loop Engineering human gate is a risk-based decision point with a named owner, a bounded question, relevant evidence, real alternatives, and…
-
Stop Rules Prevent Infinite, Confident Rework
“Continue until it works” gives an agent no stable definition of works. Stop rules convert persistence into a controlled engineering process.
-
paqad-ai AI-BOM: what this emerging product artifact records and omits
AI bill of materials software records structured facts about AI involvement, such as providers, models, workflow stages, artifacts, decisions, checks, and timestamps.…
-
Loop Engineering: Memory Without Invalidation Repeats Old Mistakes
A loop that forgets repeats discovery. A loop that never invalidates memory repeats yesterday’s answer after its assumptions have expired.
-
From prompt to evidence: inside an append-only feature ledger
An AI code audit trail should chronologically retain the feature request, route, plan and specification versions, decision records, changed scope, executed checks,…