You On AI's own evidence makes the current distribution visible. Claude Code's run-rate revenue crossed $2.5 billion with a growth curve steeper than any developer tool in history. The revenue flows to Anthropic and its investors. The productivity gains flow to the employers who deploy the tools. The workers experience expanded capability and intensified expectations simultaneously — they can do more, and they are expected to do more, and the gap between what they produce and what they are paid widens with each iteration.
The distributional question is not a matter of individual executive virtue. The boardroom arithmetic surfaces every quarter: five people can do the work of one hundred, and the market rewards headcount reduction rather than expanded ambition. The executive who chooses to keep the team, as Edo Segal describes, exercises personal moral authority against structural pressure. The choice is admirable and fragile. The distribution problem cannot be resolved by individual moral choices, however sincere, because the market rewards efficiency more reliably than it rewards moral commitment.
Gorz's proposed mechanism for redirecting the surplus was specific: taxation calibrated to the gap between augmented and unaugmented labor productivity, capturing a share of the surplus proportional to the contribution of the collective knowledge base — the training data, accumulated science, cultural and intellectual heritage — that makes the AI tools possible. The revenue would finance a guaranteed income providing material foundation for autonomous activity.
The urgency is intensified by speed. Previous technological revolutions unfolded over decades, providing time — however inadequate — for political movements to organize. The AI transition is compressing this timeline catastrophically. The twenty-fold multiplier documented in early 2026 has not yet met institutional responses adequate to its scale. Every year of delay is a year in which the surplus is captured and the concentration deepens.
Gorz developed his analysis of technological surplus through observing the introduction of microelectronic automation into European factories in the 1970s and 1980s, where identical technologies produced divergent social outcomes — Scandinavian workers capturing significant gains, others being displaced — according to the political structures within which the technology was deployed.
Three paths for the surplus. Reduced hours at maintained wages, increased profits, or displacement — and the market naturally chooses the latter two.
The arithmetic is structural. Five people doing the work of a hundred means the surplus must be distributed; the question is who captures it.
Knowledge is collectively produced. AI models were trained on accumulated human knowledge that belongs, morally, to the commons — making the surplus a social product that has been privately appropriated.
Speed intensifies concentration. The compressed timeline of AI adoption leaves no time for political redistribution unless institutions move at technology's pace.
The dam is political. Individual executive virtue cannot substitute for institutional structures that redirect the surplus toward workers.