You May Have the Wrong Mental Model for CD

“I don’t need CD. My customers don’t need changes that quickly.”

I mean, when there’s a zero-day attack happening, you may change your mind. However, CD isn’t about speed.

CD systematically removes obstacles that prevent code from reaching production — manual handoffs, long queues, environment bottlenecks — and replaces them with automated validations that catch quality issues as early as possible. Each failure reveals systemic causes, which we address by adding better guardrails, making the system learn and improve over time. The goal is enhanced quality and reduced risk. Speed is a side effect: the same small batches and fast feedback loops that improve quality also happen to make delivery faster.

CD isn’t about speed in the same way that keeping a clean kitchen isn’t about cooking faster — it’s about food safety and not sending someone to the hospital. Speed is just what happens when you stop working around the mess.

People often limit their mental model of CD pipeline validations to “developer unit tests.” They need to broaden their view of the problems. I assembled a list of where classes of defects frequently occur, where we can detect them with standard tooling, and how AI can help us move detection left in the flow for some of them or eliminate the need for manual detection of things that normal tools can’t find.