Main idea

The gap an organization tolerates between a beginner and an expert isn't absolute: it is measured against the going throughput. As long as an experienced developer and a junior produced in the same order of magnitude, letting the second spend a day on a problem the first settled in two hours was a discreet cost, absorbed by the schedule.

When a tooled-up senior produces ten times more, that same gap becomes visible and quantifiable. Accepting that a junior goes ten times slower because they are learning, or that someone spends a day on a problem an assistant bypasses in a few minutes, becomes an expense that has to be justified line by line — while the market pushes for faster delivery.

Learning time hasn't grown; it is its relative price that has jumped, because everything around it has accelerated and it alone cannot.

Why it matters

This explains why training shrinks without any decision having abolished it: it is enough for each local arbitration to favor the fast route for the cycle that reproduces the profession to break down by accumulation.

It also points to the counter-measure: learning time has to be ring-fenced as a separate line, because it loses every comparison when set against accelerated production.

Nuances and limits

The argument only holds if the company needs to manufacture its own seniors. An organization that recruits already-trained profiles externalizes the cost — and the reasoning then moves up to the level of the sector, which cannot recruit from outside itself.

And the acceleration also funds what it threatens: the time saved on production could pay for learning, if someone explicitly decides to allocate it there.

Open questions

• What form of commitment makes a training budget immune to comparison with the team's production throughput?