Beyond the Chatbox: The Rise of the Sovereign Engineer
Working code is becoming cheap. What stays scarce is judgment about what to build and whether it was worth building. That is the engineer worth becoming.
For most of my career, engineers were measured by how fast and how well they could implement. Know the framework. Optimise the query. Debug the production incident. Architect the system. Those skills still matter, but they have stopped being where the value sits.
AI has changed the economics of building software. Working code is becoming abundant and close to free. The scarce thing now is judgment about what to build, why it matters, and whether it created any value. I have started calling the person who owns that judgment a sovereign engineer: someone accountable for the outcome, from “should we build this at all” through “did it actually work.”
The bottleneck moved
Output used to track effort, and implementation ate most of the effort: gathering requirements, design discussions, writing the thing, testing it, documenting it, shipping it. A model now handles a real share of that work, including code, tests, SQL, infrastructure definitions, and the glue between systems. So the constraint moves somewhere else.
In the teams I work with, the hard problems were never the CRUD screen or the API shape. They are prioritisation, understanding the customer, deciding what not to build, and getting people to actually adopt what you shipped. Those are human problems, and cheap code has made them more exposed, not less.
Implementation is becoming a commodity
This is uncomfortable, because the profession has always rewarded technical mastery. Knowing a language or platform cold used to be a real moat. That moat is draining. When a model can produce a React component, a query, an endpoint, and a passable test suite, the value of producing those artefacts falls with it.
Engineering doesn’t matter less; the source of its value moves. Building a feature is useful work. Knowing which feature to build, validating it with real customers, and getting it adopted is the work nobody can hand to a model yet.
The people who already worked this way
The highest-leverage engineers I have worked with were rarely the strongest coders in the room. They were the ones who took ownership without being asked. They wanted to know why we were building something, what problem it solved and for whom, how we would know it worked, whether there was a simpler path, and whether we should build it at all. They behaved like people accountable for a business result rather than authors of a pull request.
AI raises the price of that instinct. When code is cheap, the decisions around it get expensive.
What I’m seeing at OptCulture
As an advisor to OptCulture, the most interesting effect of AI has not been the code it writes. It is the friction it removes. Work that used to cost an engineer several days can now be prototyped, tested, and thrown away in an afternoon.
That sounds like pure upside until you notice the new problem. When building is cheap, you can chase far more ideas than you can sensibly choose between. The limit stops being how much you can build and becomes how well you can decide. Teams start winning on customer understanding and prioritisation discipline rather than raw throughput.
The skill stack that still pays
Technical depth is the floor now, not the differentiator. The engineers I would bet on are stacking a few things on top of it: product sense, so they pick problems worth solving; systems thinking, so they see the second-order effects of a change; the ability to explain a trade-off to a non-technical room; comfort making decisions with incomplete information; and the habit of treating AI as a power tool rather than a threat. None of this is new. AI has moved it from “nice to have” to the main event.
We have done this before
Calculators did not end mathematics. Spreadsheets did not end accounting. CAD did not end engineering. The tools changed, expectations rose, and the work moved up a level. Software is making the same move. As the routine parts of implementation get cheaper, the premium on judgment, taste, and ownership goes up.
So I don’t spend much energy on whether AI replaces engineers. The better question is which engineers it makes more valuable, and the answer looks a lot like the people who already think like owners. If your edge is how fast you write code, that edge is shrinking. If your edge is knowing what is worth building and seeing it through, this is a good decade to be you.