I remember when the most comfortable place for a Software Engineer was hidden in a cubicle.

We had a system that felt safe. We stayed in the code — focused, thinking about performance, clean code, cyclomatic complexity — while Product Managers and Solution Architects stood on the front lines with customers. They translated the messy language of domain problems into technical requirements we craved, so we could refine them and design the perfect solution.

And we hated changes. Like changing the width of a bridge after the foundations were laid.

We hid in our forts. And for good reason: writing good software — well-engineered and built to last — demanded stability of requirements. We indulged in languages, platforms, design patterns, and imagination, keeping us hidden and anchored to our desks.

With AI-assisted coding, we are more agile, nimble, and insightful. We can attempt to make changes faster and reach for that 10x promise. But the real story is happening far from the velocity dream. It is the transformation of our identity.

Lately, my team has been living in the gap between the engine and the business problem — unshielded by PM or TPM, in messy fields of business terminology. With AI, we can conclude on the spot whether an idea is solvable. From there, driving the solution to completion can be a one-person job, with review from others.

It used to be stressful to inhabit that place. But we are learning quickly that it is no longer optional. Our focus is shifting from the mechanics of code toward the higher-level domain of the problem, letting agents translate intent into code while we focus on output and quality pipelines.

I felt this shift last week. I had to work on an internal codebase: a decade-old, complex system that for years was considered the territory of developers with a decade of exposure to it. With AI, I navigated that complexity and contributed to a system that used to feel off-limits.

It was a humbling realisation of what is possible: taking a customer request, understanding its value, then implementing it.

• Ask the right questions — both to the human and to the machine.
• Direct the AI toward the right solution, not just any solution.
• Have Agent/Claude.md and skills to drive the agent in the right technical direction.
• Communicate with the customer when you must make compromises.

Software Engineering remains our core competency, but our role has evolved. We must still be engineers. We need hard-earned experience to understand our choices, because only then can we guide technology toward the right outcome. That wisdom comes from mistakes. But our tools have changed, giving us a chance to grow and have more impact.

We no longer need to hide in our cubicles. We have the potential to be the bridge.

There is still work to ensure this bridge is safe in any condition the world throws at it. These changes introduce uncertainty and domain knowledge gaps, and we do not have 100% certainty this is the direction — but it is absolutely worth exploring.