Governance

Who decides, by what right, and fast enough for the decision to matter?

Built from wreckage

Many of the defining institutions of the post-war order were built from debris. The United Nations from a war that killed eighty million. The Bretton Woods system from a depression that brought democracies to their knees. The Non-Proliferation Treaty from the arms-control turn that followed the week the species nearly ended itself over Cuba. The project’s Diagnosis states the pattern coldly: for over a century it has been catastrophe first, construction second — “and the construction has never once preceded the catastrophe.”

That pattern was survivable when catastrophes were survivable. The Diagnosis’s core finding is that several of the current ones are not — not because they are large, but because they are irreversible: an ice sheet destabilized, a rainforest tipped, do not come back on any human timescale. A civilization that only builds governance after the wreck has, this time, a wreck it cannot rebuild from.

Hence this pillar’s mandate, taken from the Diagnosis’s verdict: the obstacle “is not an inherent incapacity of collective action. It is the absence, so far, of governance architectures adequate to what must now be governed.” Note what that sentence claims: that the failure is at least partly architectural — and architecture can be worked on. That is an interpretation, and a hopeful one. The strongest objection to this pillar says the failure is interest, not architecture, and this page returns to it rather than burying it.

The mismatch of clocks

The Diagnosis’s emblem makes the inadequacy concrete: to hold the least-bad climate trajectory, global emissions would need to fall further in ten years than any major economy has managed in any ten — and the institutional mechanism responsible for coordinating this convenes once a year. The response latency of the controller exceeds the time constant of the thing controlled.

The landing page compresses the pillar’s ambition into one line: deliberation at the speed of the forces it must govern. Both nouns are load-bearing. Deliberation — decisions that skip the forming of collective judgment lose the legitimacy that made them worth having. Speed — a perfectly deliberated decision that arrives after the tipping point is a eulogy.

The status of that ambition, stated at once: no mechanism yet exists, in this project or elsewhere, that delivers both at planetary scale. The pillar’s headline is its research program. What the pillar does have is a position — three commitments that organize everything it has specified and everything it still owes — and, at the end of them, a named research direction with a working definition. Both follow.

The wrong axis

The reflexive debate — centralize or decentralize? — is one this pillar declines, and the grounds now carry their sources. Hierarchies fail by bottleneck, and the classic diagnosis is Hayek’s: the knowledge a decision needs most — “of the particular circumstances of time and place” — lives at the edges and does not survive the trip to the top. Distributed systems fail by fragmentation: without shared direction, a thousand local optima add up to drift — that half is the project’s own observation, stated as such. And the strongest empirical case against the axis itself is Elinor Ostrom’s: her Nobel lecture assembled decades of fieldwork showing that the arrangements which actually govern shared resources well are polycentric — many centers of decision, nested and overlapping, neither one tower nor a thousand islands.

Both failures being real means the axis is the wrong control. The useful design variables are finer: which decisions live at which level, changeable at what speed, revocable by whom. Two principles organize the variables. Subsidiarity with federation — decide as locally as the problem allows, federate upward only what genuinely spans localities — is the Architecture pillar’s second element; what Governance inherits is its unfinished part: the map of which decision types belong at which level is itself a power question, and no one has drawn it. Requisite variety is W. Ross Ashby’s law from 1956 cybernetics: a regulator needs at least as many kinds of response as its system has kinds of disturbance. Applied to institutions it is an analogy, but a disciplined one — one assembly, one procedure, one speed cannot steer a complex system; a governance system needs a toolbox, matched rather than maximal.

Variables and principles still need a spine. The pillar’s is three commitments.

I. The governed shape the questions

Counting preferences is not forming a judgment. A vote on someone else’s framing decides only what the framer left open — which is why E. E. Schattschneider called the definition of the alternatives “the supreme instrument of power” in 1960. Question-framing is where power hides, and the project’s earlier page said it plainest, in the line that remains this pillar’s normative heart: “Citizens must be able to shape the questions, not just answer them.” Legitimacy, on this reading, is made in the deliberation, not counted in the tally.

This commitment has an evidence base the project has not previously drawn on, and the evidence teaches in both directions. The OECD has catalogued close to three hundred representative deliberative processes — assemblies, juries, panels of randomly selected citizens — across thirty-four countries, a quiet wave building since the 1980s. The strongest single case is Ireland’s Citizens’ Assembly: ninety-nine citizens drawn by lot deliberated the constitutional question elected politicians had avoided for a generation, recommended change, and framed the terms of the referendum the country then passed by two-thirds in 2018. Sortition and deliberation shaped the question; the whole polity answered it. The cautionary case is France’s Convention Citoyenne pour le Climat: one hundred and fifty citizens produced 149 climate proposals, convened under a presidential promise that the proposals would go forward “without filter” — and after the filtering that followed anyway, the citizens themselves graded the government’s follow-through at 3.3 out of 10. Deliberation without a binding interface to power is advice. And the digital frontier is real but bounded: Taiwan’s vTaiwan process used open-source consensus-mapping software to find the points of rough consensus in an Uber-versus-taxi standoff, which the government then carried into regulation — with participation in the thousands, not millions, and an advisory mandate throughout.

So the machinery exists, and has been evaluated — elsewhere, at national scale and below, with failure modes now known by name: the uptake gap, the advisory ceiling, the scale ceiling. What exists nowhere is the planetary version. And the project itself has so far adopted, proposed, and evaluated none of this machinery for its own governance; that survey is owed, and it is listed with the open problems below.

II. Power is loaned, never owned

Principles are cheap. What distinguishes this pillar from an essay is that four of its mechanisms are already specified — normatively, mostly testably — in RFC-0001, the Protocol’s draft infrastructure specification. Their scope is precise and must be stated: they bind the Protocol’s own governance, not nations. They are a working miniature, and whether the miniature scales is an open hypothesis. Three of the four rules define this commitment. The fourth belongs to the next.

Rules change at two speeds. All rule changes happen through in-protocol process. A small constitutional core — the infrastructure’s exit, plurality, anti-accumulation, legibility and power-decay guarantees — can be amended only by supermajority plus a time delay. Everything else is deliberately easy to change: easy to fork, hard to quietly capture the core. The design’s exposed nerve: whoever adjudicates what counts as “core” holds real power, and the adjudicator is not yet specified.

Power decays on a schedule. Every privileged role, key, mandate and governance weight expires on a published date and is renewable only by fresh consent of the governed. No permanent seats, no perpetual keys, no grandfathered weight. The RFC calls it power demurrage, and its test is almost comically auditable: a public register of privileged roles with zero undated entries. Consent, in this design, is a flow, not a stock. The known failure mode is renewal theatre — rubber-stamp rollovers that keep the dates fresh and the power stale — and the design’s answer, renewal as a genuine fresh decision, is easier to require than to verify.

Boundaries, monitored, with graduated sanctions. Drawing on Elinor Ostrom’s findings about commons that survive centuries: defined membership, monitoring suited to local conditions, sanctions that escalate step by step rather than jumping to expulsion, and conflict resolution cheap enough that people actually use it — with sanctions contestable and logged, and dispute latency and cost published. The specification also names this rule’s sharpest internal tension itself: credible exit requires identity people can take with them; enforceable sanctions require identity that cannot simply be shed and reborn — and a fully portable identity can outrun its reputation. No source resolves the balance.

Composed — with the emergency rule below as the fourth — the rules encode this commitment as engineering: power held on a register, decaying by default, exercised inside pre-agreed bounds, contestable at every step.

III. Speed is engineered, never seized

The two-clock problem has a tempting answer with a long body count: when the crisis comes, suspend the process. The pillar’s working stance — a stance, not a solved problem — is the opposite. Consent comes first; speed is sought through design — pre-authorization and delegation — never by bypassing consent “temporarily,” because temporary bypasses are how emergency capture begins.

That stance is a summary of the historical record, and the record teaches in both directions. The Roman Republic kept a bounded emergency office for roughly three centuries: the dictatorship was a pre-defined magistracy with a six-month automatic sunset, repeatedly used and repeatedly laid down. Whatever else preserved the republic through those centuries, the bounds were part of the design — and when the office returned after a long lapse, it returned without them: Sulla took it without the term, Caesar took it in perpetuity, and within a generation the republic was gone. The modern emblem runs the other way from the start: the Weimar constitution’s emergency article was open-ended, and the Reichstag Fire Decree of February 1933 suspended civil liberties as a formally temporary measure that remained in force until 1945, becoming a legal foundation of the dictatorship. When RFC-0001 writes that “Emergencies are how most captures happen legally; an undefined emergency power is a capture vector, not a safety net,” it is compressing this history into a requirement.

The requirement is the fourth specified rule. Emergencies are governed in advance: any emergency mechanism must be pre-defined in scope, time-boxed with automatic sunset, exercised only in public, and followed by mandatory review. You decide in calm what crisis may do, because deciding during the crisis means the crisis decides. The residual gap is real and stated: a genuine emergency outside every pre-defined scope has no answer yet, in any source.

Beyond emergencies, the speed this pillar seeks is structural, and the landing page gives the direction its name — asymmetric decentralized governance — to which the project attaches a working definition: governance that assigns different classes of decisions to different levels, speeds, thresholds, delegation rules, and revocation paths, rather than forcing every decision through one uniform procedure. The definition is deliberately modest. It generalizes what the draft specification already does at protocol scope — a slow constitutional core behind a fast periphery — and what subsidiarity, federation, requisite variety and expiring power each imply. It names a research direction with a defined target, not a solved political mechanism.

The three commitments close into a loop, as this project’s pillars tend to. Question-shaping without expiring power is captured at the framing stage: whoever runs the process entrenches. Expiring power without engineered speed arrives after the tipping point: perfect consent, delivered as a eulogy. Engineered speed without question-shaping is command, not governance: fast, and illegitimate precisely where it needs to bind. Each commitment covers the failure mode of the other two; none stands alone.

What worked, once

Skeptics of planetary governance deserve their strongest data point, and the Diagnosis supplies it: the ozone layer is healing — measurably, with ninety-nine percent of the controlled substances phased out and recovery projected within decades. The Montreal Protocol stands as the clearest case on record of a planetary-scale environmental threat being reversed by coordinated agreement. The Diagnosis also names why it worked — an unambiguous scientific signal, few identifiable industrial actors, substitutes on the shelf, a binding treaty with enforcement — and concedes that no current crisis reproduces those conditions.

That analysis cuts two ways, and the pillar accepts both edges. Against despair: collective action at planetary scale has happened, within living memory. Against complacency: the conditions that made it possible must now be manufactured — clearer signals (Architecture’s legibility), fewer effective veto points (subsidiarity’s allocation), substitutes (Economics’ work), enforcement that survives defection (boundaries and sanctions). Governance adequacy, on this reading, is not a treaty away. It is a build.

Here the entrenched-interest objection returns, as promised. The Diagnosis’s own material shows actors with full knowledge choosing extraction. If the decisive players profit from the status quo, what does better decision architecture change? The pillar’s answer is deliberately narrow. It does not claim architecture dissolves interest. It claims that interest aligned with action when specific conditions held, and that conditions are buildable — and at protocol scope, narrower still: the mechanisms above do not assume good actors; they assume adversaries, and price their attacks. Whether condition-building scales to crises with many actors and no substitutes is open, and the project says so.

What this pillar has not solved

The list is long, and it is stated here rather than discovered later. The Protocol has not solved: planetary-scale deliberation — the external machinery above stops at national scale, and the project has adopted none of it even there; representation and participation; minority protection — a conspicuous owed chapter for a project whose manifesto honours the dispossessed; intergenerational representation for century-scale aims under expiring mandates; the tension between speed and consent, beyond the working stance above; political adoption — why incumbent decision-holders would adopt rules that decay their power; the genesis problem — the founder is the first insider adversary, and no binding handover mechanism is known, with two structural checks bounding the damage meanwhile (founder roles expire like all others, and communities can fork with full state without permission); and the calibration numbers that are the design’s actual security parameters — supermajority thresholds, delay durations, decay schedules, renewal quorums.

One unwritten chapter deserves its own paragraph, because the landing page has already made a promise against it: “AI as an assistant to all of it. Never its owner.” As a governance matter this raises the pillar’s hardest question. An assistant that summarizes a million-voice deliberation, drafts the options, or synthesizes “what the community thinks” is doing the question-shaping this pillar says must belong to citizens — and it is typically provided by external model vendors, the least governable dependency in the stack. The pillar’s own principles imply what a safe design would require — plural models, auditable synthesis, contestable framings, human authority to refuse — but implication is not specification, and no specification exists. Until one does, the slogan is a binding constraint waiting for its mechanism.

How this pillar connects

Governance runs inside the Architecture (rule change happens in-protocol, on tamper-evident process) and governs it (the constitutional core is amendable only through the heavy path). It legitimately sets every economic parameter, and Economics funds it — with the unresolved tension that compensation for governance roles must not become purchasable seats. Security enforces what Governance legitimates and monitors the governance process itself for capture. Horizon supplies the criteria that make “fast enough to matter” meaningful — fast toward what — and Governance returns revisability: the process by which the destination stays chosen rather than inherited.

Where to contribute

If governance is your field — political theory, mechanism design, deliberative practice, institutional history — this pillar has more genuinely open problems per page than any other in the Protocol except its own economics. The most tractable contribution surface is calibration: thresholds, delays, decay schedules, quorums. The most urgent is the AI-in-deliberation chapter. The most scoped is the deliberation-mechanism survey: the external evidence base is named above, and what the Protocol needs is an assessment of which of it, if any, survives contact with the pillar’s own requirements. The specification’s comment process is open, and its standard applies here too: substantive objections should propose a testable alternative.

Many of the institutions we inherited were built from wreckage. The wager of this pillar is narrow, falsifiable, and worth someone’s career: that this once, the building can come first.