RACE Programming
Theorem 4 of 6 · AI-First Theory

The Paradox of Enterprise Development Cost

Theorem T4
AI exponentially amplifies engineers and displaces pure coders, raising production capacity. But demand for enterprise software, driven by vibe-coders and an insatiable appetite for automation, outpaces that capacity. Efficiency rises; cost paradoxically does not fall. It keeps growing.

Not a claim that AI makes software cheaper. The paradox is the opposite: per-engineer efficiency rises sharply, yet total enterprise development cost keeps climbing, because demand for software outpaces the new capacity.

Implication

By Pareto math, of today's ~20% engineers, the 4% who become 10x engineers cover ~40% of current capacity and the 16% in effective micro-teams add ~48%, together ~88% of today's production volume, with upskilled coders adding more. But capacity growth cannot meet the avalanche of demand. The result is a shortage and a paradoxical rise in the cost of enterprise development, which is exactly why a delivery model that maximizes engineer leverage is an economic necessity, not an optimization.

Worked example

Take the numbers in the statement. Of the ~20% of engineers who are the productive core, the 4% who become 10x one-person orchestras cover about 40% of today's capacity (4% × 10). The next 16%, in ~3x micro-teams, add about 48% (16% × 3). Together that is ~88% of current output from a fifth of the workforce, before upskilled coders. Capacity rises sharply. The paradox is that it still is not enough: demand grows faster than even an 88%-from-20% workforce can supply.

Why it holds

This is the software instance of a well-documented economic pattern, the Jevons paradox. In 1865, William Stanley Jevons observed that as steam engines grew more efficient, Britain's coal use did not fall, it rose: cheaper energy induced far more of it. The same shape recurs wherever efficiency meets elastic demand, from fuel-efficient cars leading to more driving to efficient data centers driving more compute.

Enterprise software demand is highly elastic. Every team that could not previously afford custom software now can; every manual process becomes a candidate for automation; vibe-coders and citizen developers widen the pool commissioning software. So when AI multiplies each engineer's output, the result is not a smaller bill for the same work, but a larger appetite the new capacity chases and does not catch. Efficiency per unit rises; total spend rises with it.

That is why RACE Programming treats maximizing engineer leverage as an economic necessity, not an optimization: in a Jevons regime, whoever gets the most output per engineer wins the capacity race, and the cost of falling behind compounds.

FAQ

If AI makes engineers more efficient, why do costs rise?

Because demand grows faster than capacity. Output per engineer rises, but the appetite for automation and software outpaces it, producing a shortage and a rising total cost.

Is this just the Jevons paradox?

It is the Jevons paradox applied to software: efficiency gains induce more demand rather than lower total cost. The theorem adds the mechanism, an elastic appetite for automation, and the consequence for how teams should be structured.

← All theorems

FAQ

Frequently asked questions

If AI makes engineers more efficient, why do costs rise?
Because demand grows faster than capacity. Output per engineer rises, but the appetite for automation and software outpaces it, producing a shortage and a rising total cost.
Is this just the Jevons paradox?
It is the Jevons paradox applied to software: efficiency gains induce more demand rather than lower total cost. The theorem adds the mechanism, an elastic appetite for automation, and the consequence for how teams should be structured.