Act 4 · The Limits and the Stakes · Station 10
Where centralization stops
Colocation, Mars rovers, long-tail cases, turbulence: the physical and statistical limits that force autonomy back to the edge.
The same café engine, three different worlds. Each of these levels is one parameter changed in the model you already played — and in each, no matter how much AI you buy, the frontier stops moving. Pick a level; the game relaunches with its twist. Watch what the Delegator ghost does.
Distance beats telemetry. Centralized stores now act several ticks late no matter how good your sensors get, because “communications between Earth and Mars impose round-trip lags of about 4.3–21 minutes, so landers and rovers must act autonomously.” The same physics shows up in markets, where “algorithmic trading benefits from colocated autonomy.” Buy the digital twin all you like — HQ is still reacting to a stale world. Delegation dominates whatever φ.
Plan week 1
0 centralized · 2 delegated“Steady week. Read the room, adjusted the menu, kept regulars happy.”
No dashboard here — their knowledge is tacit. You see only the result.
“Steady week. Read the room, adjusted the menu, kept regulars happy.”
No dashboard here — their knowledge is tacit. You see only the result.
AI upgrades
budget $0Monthly coarse reports
Analyst (K̄=1) — refresh 1/week
Where economics runs out
The three levels above are economic limits — physics, statistics, turbulence. But Brynjolfsson & Hitzig are careful that “the ‘ownership regimes’ of organizations … are, of course, not made on the basis of economics alone” (Brynjolfsson & Hitzig, 2025, §6.3). Some decisions stay at the edge because law or legitimacy keeps them there. These are not levels — you cannot buy your way past them.
“Professional licensing (e.g. for pilots, physicians, engineers) impose human-in-the-loop obligations that could, if upheld, continue to distribute decision rights” (Brynjolfsson & Hitzig, 2025, §6.3). A perfect central plan is worth nothing if the law requires a licensed human to sign it.
“Even without mandates, people may continue to want humans in control of consequential or meaning-laden decisions: a judge for sentencing, a pilot during critical phases, players on the field, or a human coach, poet, or therapist.” These preferences “sustain local agency where legitimacy, dignity, or narrative matter” (Brynjolfsson & Hitzig, 2025, §6.3).
“Sectoral laws in finance, health, transport, and criminal justice may continue to require human oversight, contestability, and auditability; that often anchors decisions where traceability is strongest, not where coordination is cheapest” (Brynjolfsson & Hitzig, 2025, §6.3).
Put together, the economic limits and the legal ones frame the last question of the arc — not whether machines can centralize decisions, but what it means when they do. That is the stakes.