There is a phrase circulating on a technology podcast — episode 269, if you require the coordinates — that has lodged in my ear and will not dissolve. The phrase is: 'governance needs to become exponential.' It was offered by four men debating whether the nation-state can regulate artificial intelligence, alongside Sam Altman's reported gesture of handing five percent of OpenAI to some public benefit, and the general dread that machine intelligence is outpacing the institutions meant to contain it. It has the grammar of an idea and the metabolism of a slogan. Governance is not a quantity that can grow exponentially, any more than justice can be scaled or deliberation compiled; the metaphor is borrowed from the machine these men are frightened of, and it smuggles in the assumption that politics is an engineering problem awaiting throughput.
Something true is trying to speak through the phrase. The speakers have muffled that truth in the only language they trust.
Let me do these men the courtesy I do the dead: let me build their argument better than they built it. Strip away the engineering metaphor and something real remains. Institutions designed for a slower world do genuinely struggle against a technology that iterates faster than a legislature can convene. That is not nonsense. Madison worried in Federalist No. 10 about factions multiplying faster than deliberation could accommodate them.
The speakers have intuited a real asymmetry. What they lack is not the intuition. What they lack is the vocabulary that would let it survive contact with history.
So let me lend them the vocabulary, and follow it where it leads. The strongest form of their claim is this: representative government is itself a technology, evolved to aggregate dispersed human judgment across time, and its slowness was never a defect but a specification — a deliberate friction that gives contested interests the interval in which to contest. What has changed is not that the friction has gone bad. What has changed is that the timescale of consequence has collapsed. A model trained over a single summer can restructure a labor market before a single hearing is scheduled.
The speakers feel this collapse, and they are right to feel it. But here their frame, extended honestly, turns against their own conclusion. If the trouble is that consequence now outruns deliberation, then abolishing deliberation in favor of something faster is not a solution — it is surrender to the very acceleration they diagnosed. The honest question is narrower and far harder: which functions of deliberation can be made to operate at the new speed, and which cannot be accelerated at all without becoming something other than deliberation — a rubber stamp, a notification, a fait accompli with a comment period. That is not an engineering question.
It is a question about what we are willing to lose.
Notice what 'exponential governance' quietly discards. It keeps the decision and throws away the deliberating. This is not a new dream. In the 1930s an outfit called Technocracy Incorporated, led by an engineer named Howard Scott, proposed to replace the whole apparatus of politics with administration by technical experts, who would run the economy by measuring energy in ergs rather than dollars, and who saw parliaments as inefficiency to be optimized away. It drew crowds.
It promised to make consequence rational. It could not survive the first question any voter would ask, which was: accountable to whom?
That is the question the phrase is built to avoid. Governance 'needs to become' exponential — the passive voice does the work. Who makes it so? By what legitimacy? Answerable to whom, when the henhouse inspector is nominated by the fox?
The acceleration these men fear is real, and their alarm is more honest than the complacency of officials who feel nothing. But they have made the technocrat's oldest error: they have found an absence — no institution quick enough for their machines — and proposed to fill it by removing the one thing that made institutions worth building. The friction they wish to optimize away is not a defect in the system. It is the system. It is the only arrangement yet devised in which a neighbor about to lose her job may say so, aloud, to someone obliged by law to hear her and answer.
Speed cannot manufacture that obligation; it can only outpace it and leave her addressing an empty chamber. And the whole seduction of exponential governance is that it lets a handful of men outrun accountability while calling the flight progress.
