Point of view
The Jevons Paradox and the coming shortage of engineers
AI will not reduce demand for software engineers. It will create a shortage of them. Here is why we believe that — and why it shapes everything we build at Unio Lab.
The paradox
In 1865, the economist William Stanley Jevons observed something counterintuitive: as steam engines became more efficient, coal consumption went up, not down. More efficient machines didn't reduce demand for the resource — they made the resource economically viable for more applications, which multiplied consumption faster than efficiency could save it.
This is the Jevons Paradox. And it applies, with striking regularity, every time a technology dramatically lowers the cost of entry to a market.
The pattern is older than software
When smartphones put a production-quality camera into every pocket, the global market for professional camera operators and editors grew. When Excel automated accounting functions, demand for accountants and analysts increased. When YouTube collapsed the economics of television — from €5M per episode to a man cleaning rusty tools in his shed with 200K subscribers — it created an enormous new market for video producers, editors, and directors.
The same logic applies every time. Chefs who publish their recipes free online find their restaurants fully booked. Consultants who give away their best ideas in books get hired at higher rates. Making something cheap doesn't destroy the market. It expands it.
Excel · 1985
Automated core accounting functions → grew demand for accountants
Smartphone cameras · 2007
Democratised photography → grew demand for professional operators
YouTube · 2005
Collapsed TV production costs → grew demand for video production talent
Software is next
What used to cost €500,000 and six months of a dedicated engineering team now costs €5,000 and two weeks. That cost collapse means the number of viable software companies isn't staying flat — it is growing by orders of magnitude. Ideas that previously needed 100,000 users to break even now need 1,000. The total addressable market for software is expanding faster than AI can displace the engineers inside it.
Each of those new companies — every one of the millions who will now build what they previously couldn't afford to — eventually faces the same problem: the codebase that worked at fifty users breaks at five hundred. The schema that made sense in week two creates a migration crisis in month eight. The architecture that shipped fast becomes the liability that prevents the next feature from shipping at all.
That's not a problem AI agents solve. That's a problem senior engineers solve. And there are now far more companies who need that expertise than there are engineers who have it.
Why this shapes how we build
At Unio Lab, we believe the value of engineering rigour doesn't fall as AI lowers implementation costs — it rises. When AI does the volume, what remains scarce is the human judgement that determines whether a system will hold together at scale.
That's why our methodology puts specification first: a precise, machine-readable definition of the system before a line of code is written. It's why our engineers have seven or more years of experience in cloud-native architecture and industrial platform engineering. And it's why we verify every AI-generated output against a contract before it ships.
We're not using AI to replace engineering. We're using it to make the parts of engineering that matter most available to the teams who need them right now — fast, rigorously, without the overhead of a twenty-person agency.
The Jevons Paradox has surprised us before. We think it's about to surprise the software industry. We built Unio Lab to be on the right side of it.
Have a system that needs engineering?
Describe the problem. We'll assess whether it's a good fit for spec-driven delivery and scope it within a week. If it's not right for our methodology, we'll tell you.