State Capacity
under AGI
An instrument for exploring how the modern state might be remade when machine intelligence becomes abundant. Choose two driving forces, and four contrasting worlds are generated at the corners of the resulting matrix — each explored through structural features, service delivery, winners and losers, and policy pathways traced ten and twenty years out.
The state is best understood as an answer to a cognitive problem. No individual mind can know a territory, anticipate a harvest failure, coordinate a million strangers, or hold a promise stable across generations. Human cognition is bounded — in attention, in memory, in the reach of trust — yet societies must sense, decide, and act at scales far beyond any single mind. The institutions we call the state are the technologies invented to close that gap: the census and the register extend perception; the bureaucracy extends memory and procedure; courts stabilise expectations; taxation and mandate convert dispersed intentions into collective action. State capacity is, at bottom, cognitive capacity — institutionalised.
AGI matters for the state because it re-prices exactly these tasks. When perception, coordination, and judgement become abundant, cheap, and available to actors other than governments, the constraints that made the state necessary begin to loosen — and the settlement those constraints produced comes open for renegotiation. Whether the state deepens, hollows, fragments, or is quietly outcompeted is not determined by the technology; it is determined by how these pressures resolve. That resolution is what this instrument explores.
Every generated world is read through the same three lenses, because they name the three cognitive constraints the state historically solved. The dots on scenario cards show intensity — from dim (the feature barely registers) to bright (dominant). Each lens can swing to either extreme.
Data disperses across platforms and individuals. No actor has privileged sight. The state must compete for data about its own population.
The state gains unprecedented informational reach via AI. Citizens become fully legible to government in ways never previously possible.
Coordination is emergent, driven by platform logic and commercial incentive. Uneven, with structural gaps where market signals are weak.
Functions bundle under unified state control. AI collapses coordination costs inward, reversing decades of delegation and outsourcing.
Legitimacy becomes plural and domain-specific. No single actor holds comprehensive authority. The state must earn relevance rather than inherit it.
State capacity expands faster than accountability can follow. Formal authority persists but democratic oversight struggles to keep pace.
One way to see the renegotiation coming is to decompose the state morphologically: break it into the functions it performs, and ask which actors can credibly perform each one. Solid marks show who holds or shares a function today; rings show who can newly contest it once advanced AI is generally available. Click a function for how AGI redraws that row.