Observability Engineering, 2nd Edition

Chapter 2. How Code Crosses Over: Validating Developer Intent in Production

Developers develop preproduction, operators operate postproduction, and deploys mark the handoff point with a switch. This groove is etched so deeply that many companies don’t even let developers look at their code in production. At other companies, it’s just so awkward and cumbersome and difficult, and the payoff so minor, that they don’t even try.

The dev/ops divide has persisted because it reflects both the need for change and the need for stability, in a way that maps neatly to org charts, budgets, and goals. But AI is collapsing the distance between writing code and understanding what it does in production—indeed, AI is forcing all the stages of the SDLC to blur and combine and be reinvented. An agent doesn’t know if it’s designing, testing, or coding. What matters is intent, context, and validation.

When code crosses over, it must be validated—not just in theory, with testing, but in reality. There are many techniques for validating code in production. These are not new. But they have never been widely adopted, because they are complicated and deploying them is hard. AI is challenging this from many sides: by making it technologically easier to achieve, by applying competitive pressure, and by upending the economics of code production.

This chapter covers the practices that make that possible. They form a flywheel: each one enables the next, and together they accelerate and amplify each other’s capabilities. ...