Engineering moves before and after the code

I know that coding agents are inevitable and the next natural step in the AI-powered world, but I’m still feeling anxious every time I set them loose on a task.

Sometimes it genuinely feels like I’m losing control over the very heart of software engineering - the act of writing code itself.

At the same time, when an agent is given all the necessary context, clear objectives, solid guardrails, and well-defined architectural decisions, the code it delivers is surely better than anything I would write myself under the same constraints.

So for me, the actual engineering has shifted to the pre-code and post-code work: architecture, defining clear objectives, agent orchestration (or “harness,” if you prefer that term), CI/CD pipelines, guardrails, and automation.

It can feel overwhelming at times - becoming hard to keep up with and quite energy-draining - but here’s the most important part: I’m still genuinely enjoying the process.