Paradoxe de Jevons

William Stanley Jevons formulated it in 1865; the mechanism it names is that gains in efficiency raise consumption rather than lower it. Four source notes reach for it, and each applies it to the same object: code whose cost of production is collapsing.

Simon Wardley, in a fictional Socratic dialogue dated 27 March 2026, uses it to answer whether LLMs and «vibe coding» mean more or fewer developers. His answer is roughly as many, because competing firms reinvest every productivity gain to hold position: a large company goes from 30 million to over a billion lines of code just to stay where it is. Philippe Ensarguet pushes the same logic to its limit: when execution cost falls to near zero, demand becomes infinite, and the expectation is not less work but a thousand times more output.

Dave Farley reads the identical mechanism as a liability. Cheap code means more code, more complexity, more integration points, more behaviours to evaluate, and probably less time spent understanding the problem. He calls it a complexity bomb with a delayed fuse.

The DG Trésor note Trésor-Éco n° 391 gives the argument its economic form: exposure to automation does not determine employment, price elasticity of demand does. Developers and graphic designers sit above 1, so falling costs can grow their employment. SFEIR calls this structurally pro-employment for developers.

What none of the sources settle is whether elasticity above 1 survives an agentic scenario, which Trésor relegated to a footnote.