Your unrelated skills were never unrelated

There is a particular kind of professional shame that has no name. It belongs to people whose careers do not summarise well — who, asked what they do, find themselves giving a two-minute answer and watching the other person wait for the short version. They have usually been told, at some point, by someone senior and well-meaning, that they need to focus.

The advice was not wrong. For most of the twentieth century it was among the most reliable predictions anyone could make about a working life. What has changed is not the wisdom of the advice but the machinery underneath it.

Why narrowing paid

Specialisation is not a cultural preference. It is an economic response to the cost of coordination. When it is expensive to move knowledge between people, the efficient move is to keep each person inside one body of knowledge and connect them through institutions — firms, professions, licensing bodies, departments.

That is what the industrial era did, and it worked spectacularly. Adam Smith opened The Wealth of Nations with a pin factory precisely because the arithmetic was so stark: ten specialists produced orders of magnitude more than ten generalists.¹ Every institution we still work inside was built on some version of that observation.

The cost of that arrangement was paid by individuals, and it was mostly invisible. If knowledge lives inside disciplines, then a person who moves between disciplines is repeatedly paying the entry fee and rarely collecting the returns. They look, from the outside, like someone who cannot commit.

What actually changed

It is tempting to say artificial intelligence changed this, and tempting because it is partly true. But the shift began two decades earlier and had very little to do with intelligence, artificial or otherwise. It had to do with access.

Consider what has happened to the cost of learning something outside your field. Lectures from the best universities became free and watchable. Documentation, tutorials and the recorded reasoning of practitioners moved online in volumes no library ever held. Software that once required a licensing negotiation became a monthly subscription, then often free. Manufacturing became reachable through a browser. Distribution stopped requiring a publisher, a broadcaster, a label or a shop.

None of that was framed as a transformation of expertise. It was framed as convenience. But collectively it did something specific: it removed the gatekeeping that made knowledge scarce, and it did so across every discipline at roughly the same time.

And then the second shift

Artificial intelligence is the newer layer, and it is different in kind rather than degree. The earlier changes lowered the cost of understanding something. This one lowers the cost of doing it.

Those were always separate problems. You could read about typography for a year and still not be able to set a page. You could understand statistical inference perfectly well and still be unable to write the code. The gap between comprehension and execution was measured in years of deliberate practice, and it was the main reason breadth stayed expensive: crossing into a new field meant paying that cost again from zero.

That gap is narrowing. Not closing — the distinction matters enormously — but narrowing enough that a person who genuinely understands what good looks like in a discipline can now produce competent work in it without having spent a decade there.

The part that does not transfer

Here is where the optimistic version of this argument usually overreaches, so it is worth being precise about what has not changed.

Knowing what good looks like is not free, and it is not something these tools provide. It is acquired the old way: by doing the work badly for a long time under someone who can tell you why it is bad. A person who has never developed taste in a domain cannot evaluate what a machine produces in that domain, which means they cannot use it well — they can only accept its output and hope.

This is why the people benefiting most visibly from these tools are not beginners. They are people with deep training in one field, using the tools to operate in a second. The depth is what makes the breadth usable.

What this means for a scattered CV

If the argument holds, several things follow that are worth stating plainly.

• A career that does not summarise neatly is not evidence of a problem. It may be evidence of an unusual asset register.
• The value is not in the number of fields, but in the distance between them. Two adjacent specialisms compound less than two distant ones.
• Range without depth in anything remains what it always was: a liability. This is not an argument for dabbling.
• The advantage is not automatic. It arrives only when someone actually connects the fields, which is work, and which most people with varied backgrounds never get around to doing.

The last point deserves emphasis, because the comfortable reading of this essay is that everyone with an interesting past is secretly well positioned. They are not. Accumulated range is potential energy. It does nothing until it is put to work against a specific problem, and the people who convert it are the ones who go looking for the problem that needs exactly their combination.

Not late

What is genuinely new is not that some people have wide-ranging backgrounds. There have always been such people, and for a century the structure of work quietly penalised them.

What is new is that the penalty is lifting. The detours were always accumulating something. Now there is finally a way to spend it.