Jean Barrère's latest book, Économie du numérique, opens on a sharp observation: technology moves in exponential regimes while the institutions meant to absorb it keep operating on linear ones. He calls the gap the "Grande Divergence" and compresses it into a single line: "technology doubles, institutions add."
The doubling side has an old, precise formulation. In 1936, Theodore Wright, an aeronautical engineer at Curtiss-Wright, counted airplanes and noticed that every time cumulative production doubled, unit cost fell by a stable proportion. The same curve later showed up in solar panels and, since 2022, in the price of AI tokens, which suggests a general mechanism: accumulated practice turning into less friction, whatever it is applied to.
The control plane
The other side is the control plane: everything meant to govern how a technology gets industrialized, from budgets voted over twelve months to studies, standards and validation steps. Each layer is reasonable on its own, and each was built for a world where the thing governed changed more slowly than the governing.
Technology doubles. Institutions add.
Fast technology reveals reaction speeds that were already there, as chromatography separates a mixture. Stacked together, the layers add the latency that Barrère's line predicts: they answer a landscape that has already moved.
Our instruments no longer measure the world. They measure our delay.
Aviation also met this problem early. Past a certain speed, a pilot can no longer move the controls by hand, whatever their training. The answer was a different control system, hydraulic boost first and fly-by-wire later: the aircraft had changed regime, so the control laws had to change with it.
Fly-by-wire kept control and relocated it, from mechanical transmission to sensors, software and fast feedback loops. Organizations facing exponential technology may need the same relocation: fewer sequential approvals, firmer boundaries, faster feedback, and more authority where the signal is produced.
Where it already happens
Such a redesign usually starts small, following a pattern that has become predictable: a team forms across the seams of the org chart around a gap nobody has been assigned, proves the case by shipping, then joins the existing structure through people already inside. Generative AI has shortened that cycle from years to months.
A few weeks ago, during a reflection session, we looked back at the path our team has travelled in recent years, marked by the disruption of generative AI, which has run through our whole organization and all our projects. We tried to name what still held us together beyond architectures and project boundaries, which have become porous. Nobody reached for the org chart. The word that came back, from more than one person, was culture: a shared appetite for exploration that compounds faster with each cycle it survives. That culture is the substrate that lets teams with real autonomy stay coherent without going back to the org chart.
The organizational analogue of Wright's curve may be this: friction falls with cumulative recomposition, just as unit cost falls with cumulative production. An institution that has never recomposed itself stays linear, whatever the pace of the technology around it. The open question is what a control plane would look like if it learned as fast as the thing it governs.
