Superagency as the Test of the Wager
Superagency, published in 2025 with Greg Beato, is Hoffman's most concentrated argument that AI expands rather than contracts human capacity. The title is a deliberate provocation. It collapses the worry that AI will make humans superfluous into the opposite claim: that AI will make humans super, not in the comic-book sense but in the sense of having access to capabilities that were previously the province of teams, institutions, or elites. A lawyer for every person. A doctor for every person. A tutor for every person. A strategist for every person. The promise is not that AI replaces humans but that it democratizes access to the kinds of help that were previously available only to the rich.
The framing depends on a specific reading of historical technology. Hoffman draws repeatedly on the analogy to literacy. Writing was once the privilege of scribes; mass literacy redistributed the cognitive infrastructure of a civilization. Computing was once the privilege of corporations and governments; personal computing redistributed it. The internet was once the privilege of academics; web browsers redistributed it. Each redistribution produced losses — scribes lost their monopoly, mainframe operators lost their priesthood — but the aggregate gains in human capacity were so large that the losses, while real, were eclipsed. Hoffman expects AI to follow this arc, and he expects the timeline to be compressed.
The argument is not naive about distribution. Hoffman acknowledges that the gains will not be evenly distributed in the first instance. The early access goes to people with capital, infrastructure, and skill. But he argues that the cost curves of AI are dropping fast enough that mass access is achievable within a decade, that the floor of capability for the average person rises faster than the ceiling for elites, and that the net effect is a compression of inequality of cognitive resources even if other inequalities persist. This is an empirical claim, and one that will be testable.
The critique from the left is that the analogy to literacy is misleading because literacy is a stable capability that, once acquired, belongs to the literate person. AI capability is rented from a model provider, can be revoked, can be priced out of reach, and can be configured to serve the provider's interests rather than the user's. Superagency, in this reading, is at risk of becoming super-dependency. Hoffman has addressed this. He argues for the importance of open models, of competitive markets among providers, and of regulatory regimes that prevent monopoly capture of the cognitive substrate. Whether these prescriptions are sufficient to prevent capture is, again, going to be adjudicated empirically.
The critique from the right is that Superagency overstates the friction between AI and existing institutions. If everyone has a lawyer in their pocket, what happens to courts? If everyone has a doctor in their pocket, what happens to medical licensing? If everyone has a teacher in their pocket, what happens to schools? Hoffman's response is that institutions will need to adapt, that some of them deserve to adapt because they have ossified, and that the disruption is the cost of the gain. Whether the adaptations happen in time, and whether they preserve the legitimate functions of those institutions, is also a real question.
Superagency, in the end, is Hoffman's bet that the AI transition can be framed as a story of expansion rather than replacement. The bet is not that nothing will be lost. It is that the gains, properly distributed and properly stewarded, will be large enough to absorb the losses. This is the most concentrated form of the techno-humanist wager. It is also the wager that the next decade will most directly test. Hoffman has staked his late career on it. The orange pill, in his version, is the willingness to stake everything on a future that is not yet arrived.
