The stress test · Steelmanned
The hardest questions.
A framework should be judged by its strongest objections, not its weakest. These are the hardest we and our critics have found. Each is stated as a fair-minded skeptic would put it, then answered as plainly as we can, conceding what we must. Several remain unresolved, and those are marked plainly at the end.
These are drawn from the framework's own objection record and red-team table, sharpened against the academic critiques of comparable systems. Where the manifesto already concedes a problem, the answer says so rather than pretending to more than it has.
01
The floor is a throne
The objection Your robustness story is that the core is held outside the loop as non-negotiable. But a clause that overrides every other and never sunsets cannot apply itself, so someone must interpret it and read the un-summable panel to set a mode. An interpreter that is permanent, general, and answerable only to the clause it interprets is, by your own P–05, a sovereign. You abolished sovereignty in every seat except the one that adjudicates all the others, and the anti-capture regress does not converge on safety, it converges on a single apex.
The answer This is the framework's new lead open problem, OQ–00, and the honest verdict is a reframe, not a rebuttal. Sovereignty cannot be dissolved here, only slowed and relocated, so we stop claiming to abolish it and claim instead to replace a permanent personal sovereign with a plural, decaying, contestable stewardship of the core. Non-derogable cores with no single sovereign interpreter do exist: jus cogens, the halakhic rule that the law is not in heaven, and a constitutional court that is plural and term-limited, which is the model. What we concede plainly is that reading the panel by its weakest sign is a judgment, so the residual sovereign is whoever frames the methodology, and that is now a named, watched capture surface rather than a solved problem.
02
The economics needs the sovereign the politics forbids
The objection Enough is humane only if a universal floor of basic services comes first, and you concede that without the floor the criticism holds. But a universal floor is non-excludable, and provision at that scale has always needed a durable authority with the power to tax and bind, which is the sovereignty you exist to abolish. Your own commons evidence, Ostrom, works precisely because commons are bounded and excludable. The economic program presupposes the state-like capacity the political program dissolves.
The answer The contradiction is real in function and resolvable only in form, and we say so. The decisive word in our definition of sovereignty is general. A single-purpose provisioning commons, financed where possible as a dividend from common wealth, held as a fiduciary trust, persisting by charter while its officeholders rotate under decaying standing, is durable without being general or self-reinforcing, so it is not the sovereign we forbid. What this does not escape is the function: the floor still needs jurisdiction-wide, compulsory finance, and whether a chartered, decaying body can hold that without re-growing into a sovereign by the Robert Moses route is an open test, now on the frontier.
03
There is no theory of the outsider
The objection Every adversary you model is an insider bound by your rules. You have no theory of the rival civilization that adopts none of this. Among defectors your virtues invert into attack surfaces: the open commons leaks your position, protected dissent is a channel a hostile state funds and your sacred core forbids you to close, and decaying authority with a peacetime drawdown leaves you out-armed. Resilience matures over decades while conquest is decided in years, so you are eaten before your advantage arrives. Aeviterna may be viable only as a tenant of someone else's security order.
The answer This is the second new open problem, OQ–08, and the verdict is partial rather than fatal. Game theory itself points to the fix, since cooperators survive among defectors only in clusters, which argues for a security commons of allied polities plus a porcupine, deterrence-by-denial posture that armed neutrals like Finland and Switzerland show is compatible with deconcentrated power. The two fixes you say we forbid are less forbidden than they look: effective intelligence can keep accountability while surrendering publicity, and a deliberative human override is already how real deterrence is run. What we cannot yet answer, and now state as the falsifier, is the timescale, since the cluster must be built before the attack and we are most exploitable while building it. The perimeter is now a worked Draft in the framework's Boundary Layer.
04
Tenant, not building
The objection Your four anchors, Ostrom's commons, Mondragón, vTaiwan, and Wikipedia, are all sub-civilizational and all run inside a substrate they did not create, a state's courts, a currency, a defended internet, a market, and each erodes at the boundary with that larger system. None is evidence the model works as the building rather than as a well-behaved tenant of one. And the defining problems of the century, climate, pandemics, nuclear weapons, AI, and financial contagion, all pull authority toward the center you abolished, so the framework is least applicable exactly where the hard problems live.
The answer This is the most damaging objection in the set, and we concede it and reframe rather than defend. The anchors are now labeled tenant-grade, evidence that the primitives work as tenants, not that the whole stands on its own, and the deepest gap, integration at the federation scale, is marked None yet. You cannot test a whole civilization, but you can test the interface, so capture propagation, contagion, and metric drift across coupled units are simulable as a partial harness, and the praxis staircase is now read as an evidence ladder where each rung pre-registers the next rung's failure. On the center-pull, the strong claim is false, since polycentric climate governance is real, but the weak claim holds, so we concede a thin central spine for the few irreducibly central functions and bind it to the same decay, audit, and exit as everything else.
05
Unfalsifiable by construction
The objection Your falsifiers only weaken or revise modules, your core is declared normative not empirical, and your modularity plus a changelog let any disconfirmation be absorbed as a local revision while the whole survives. The load-bearing claims are the least falsifiable: rigid systems have lasted centuries, so P–01 is rescued by latent brittleness that has simply not arrived, and P–02's without active resistance is unfalsifiable both ways. A framework that cannot be killed by evidence, only fed by it, has dressed a vice as a virtue.
The answer Fair, and we have changed the methodology rather than argue the point. The ledger is now split: the value axioms, dignity and non-domination, are normative and not up for empirical refutation, but the causal claims, P–01 and P–02 and the slogan, are wagers that inherit empirical risk. We added a stopping rule, so a core claim that fails its test is withdrawn rather than patched, and we named the would-be refuters in advance, the durable extractive regime for P–01 and the uncaptured low-resistance system for P–02. The banner, equilibrium is death, is now cashed into its testable form: suppressing a system's feedback and adaptive capacity raises its collapse hazard conditional on shock, which the record supports better than the slogan, since the longest-lived orders survived by adapting rather than holding still.
06
The consent paradox
The objection A framework that fixes the vital signs and a dignity floor in advance, and puts some values beyond a vote, binds people who never agreed to it. Calling that voluntary through a right of exit means little when there is nowhere uncaptured to exit to, and a system built for civilizational scale shrinks the outside.
The answer We own this as an irreducible tension. The framework rests legitimacy on low-cost exit, voluntary adoption, and a fairness test taken behind a veil of ignorance, rather than on the fiction of tacit consent. What it cannot escape is that the dignity floor is asserted as non-negotiable rather than agreed, modeled on the inviolable-dignity clauses of constitutional law. We choose a small entrenched core over pure proceduralism, and we accept the legitimacy cost instead of pretending it away.
07
The transition problem
The objection Anyone who profits from extraction will use their power to block a system designed to limit it, so there is no incentive-compatible path from here to there that avoids either a destabilizing crisis or a coercive vanguard, which recreates the concentration the framework opposes. Prefigurative politics, its chosen vehicle, has a long record of staying marginal.
The answer This is the framework's first standing open problem, and it claims no solution. The proposed path is a two-stroke engine: build parallel institutions during normal times, then adopt them opportunistically when the incumbent system hits a legitimacy cascade. The honest concession is that this depends on a forcing function we can neither manufacture nor schedule. Aeviterna has a theory of readiness, not a theory of victory.
08
Who watches the watchers
The objection Whoever sets the thresholds and defines the vital signs becomes the new sovereign, because the power to define health is the power to define dissent. Adding oversight of the oversight only moves the problem, since there is always a final, unaudited monitor.
The answer The framework concedes that recursive oversight of the monitor is named and left unsolved. Its defense is to apply the full anti-capture kit to the monitor itself: plural competing measurers, panels rotated partly by lot, open data, and definitions that expire so a frozen taxonomy cannot quietly rule heterodoxy out of existence. That makes capture harder and more visible. It does not make capture impossible, and we do not claim otherwise.
09
Goodhart's law
The objection The whole system runs on measured thresholds, which makes it maximally exposed to Goodhart's law: once a measure becomes a target, it stops being a good measure. A civilization metering trust, ecology, and cognition is metering exactly the fuzzy goods most easily gamed.
The answer We state this against ourselves in three places. The mitigations are concrete: rotate indicators, keep some of them qualitative, decouple measurement from the people it rewards, and watch a band rather than maximize a number, treating metrics as evidence for judgment rather than as an autopilot. The concession that remains is that gaming-resistance is mitigation, not immunity, and the reliance on measurement is structural rather than removable.
10
The iron law of oligarchy
The objection Michels showed that any organization complex enough to need coordination grows a self-entrenching elite. Term limits and decaying standing do not defeat that. They convert formal office into informal influence as the same insiders recirculate through new seats and the bodies that write the rules.
The answer The framework calls this its deepest structural risk and concedes that no mechanism fully closes the gap. It tracks network concentration rather than individual turnover, pays bounties for exposing capture, applies decay to administrators too, and fills some roles by lot. This is containment of a permanent tendency, not a cure. Naming the adversary and rewarding its exposure raises the cost of recirculation without abolishing it.
11
The scale problem
The objection Deliberative and polycentric governance work at the scale of a community or a single institution. The record shows participation thinning and capture rising as you scale up, and indivisible global problems pull authority back toward a centre, the very thing the framework exists to prevent.
The answer This is an open problem, and we concede the residual: some global problems do pull toward the centre. The reply is to scale fractally, through nested units coordinated by thin shared protocols rather than one platform, with exit as the deepest check. Ostrom herself cautioned that her principles came from bounded physical resources and that their transfer to a global digital commons is unsettled. The scale claim is a hypothesis modeled on commons governance, not a demonstrated result.
12
Amateur government
The objection Sortition, rotation, and decaying standing strip out the expertise and institutional memory that complex governance needs. If the most capable are least willing to serve short rotating terms, the body ends up less competent than the population it governs, and the permanent bureaucracy holds the real power.
The answer Here the framework does not yet have a full answer, which is why the operation of standing is an explicit open problem. The direction is a per-domain reputation that decays on a tuned half-life, so deep expertise in slow-moving fields persists longer than in fast-moving ones, backed by redundant archives and restoration drills that hold memory outside any one tenure. The half-lives are uncalibrated, and the bet that tuned decay keeps the anti-capture benefit without paying rotation's full competence cost is unproven.
13
The panopticon
The objection A system built on transparency and continuous monitoring rebuilds the panopticon. People who know they may be watched censor themselves, and the harm ratchets one way, since each new act of measurement brings an immediate benefit while the cost to dissent is diffuse and deferred.
The answer We concede the ratchet by name. The countermeasures are a hard coherence ceiling that forbids total observability, protected ambiguity zones with a no-measurement default and wardens accountable to the people inside them rather than to the optimizer, and methods that verify a process without exposing its content. The honest point is that these zones must be permanently entrenched against a standing pressure to erode them, and whether that holds over decades is untested.
14
Degrowth in disguise
The objection The economics of enough is degrowth relabeled, and degrowth has an incentive problem: pulling back from growth causes hardship, especially for those with least. If green growth can cut emissions without contraction, then enough is not just painful but unnecessary.
The answer The framework gives a conditional answer and concedes the condition. Enough means securing a floor of basic provision first, so sufficiency is reached without contracting what people need, and it grounds the idea thermodynamically as a band between an energy floor and an ecological ceiling rather than as undifferentiated shrinkage. The whole argument leans on first delivering that floor, itself a vast unsolved project, and if rapid decoupling proves achievable the case for enough weakens.
15
A single point of capture
The objection A central knowledge commons is one high-value target. Whoever captures its governance controls the shared epistemic substrate of the whole system, and peer production has a documented gap between its egalitarian image and its drift toward informal oligarchy.
The answer The defense is to refuse a single point. The commons is governed by Ostrom's principles at scale, with clear boundaries, monitoring, graduated sanctions, and nested units so failures stay local, plus a right to fork. We concede that knowledge is not a pasture, that the transfer of commons theory to a global digital commons is unsettled, and that the capture risk of peer production is real. The answer to single-point capture is to deny that there is a single point.
16
Emergency powers and manufactured crises
The objection Every system with an emergency mode eventually has it abused, because the actor who benefits from crisis powers is the one positioned to manufacture the crisis. A broadly popular manufactured crisis is the worst case, since it clears every check legitimately, and automated thresholds add a new attack surface.
The answer The framework names the residual it cannot beat: a broadly popular manufactured crisis can still clear these checks. The defenses are specific. Separate the one who declares an emergency from the one who benefits, make emergency powers default to expiry with a heavy re-authorization burden, let nested units veto, decay emergency standing fastest, and use externally audited triggers with a deliberative override. Procedure can stop the unpopular grab. It cannot stop a truly popular one.
17
No organ for the machines
The objection You call yourself a design for an AI-mediated civilization, yet you have no treatment of advanced AI. It is the one force that accelerates both of your named enemies at once: it cheapens the surveillance and persuasion that capture feedback, it removes the many-cooperating-humans constraint that made broad sovereignty expensive, and it is already moving inside the loop that reads your vital signs. A fast or quiet AI transition is the exact disturbance your adaptive stability was built to survive, and you say almost nothing about it.
The answer Conceded, and now named as OQ–09. The direction is the framework's own logic pointed at the machine: an AI may be a monitored actor holding scoped, revocable, decaying standing, never a sovereign; consequential automated decisions carry a right to human review; the systems that read the vital signs must be plural rather than a monoculture; and frontier systems inherit the reversibility discipline of EQ–05c. The honest residual is sharp, since a gradual handover of human judgment could erode the very feedback channels the framework reads and defeat its own detector, and the worked organ is held in draft rather than live. The full attempt is in the companion paper, The Contested Mediator.
18
Who speaks for the river
The objection Every protected interest in your framework gets an agent and a decaying-standing account, except the one that underwrites all the others. Nature appears only as a ceiling on throughput and a dial on a panel, never as a party with standing of its own. And the moment you try to fix that by giving an ecosystem a guardian, you rebuild your deepest problem: whoever speaks for a forest can become its unaccountable sovereign.
The answer Conceded, and now named as OQ–10. The direction reuses the guardian-of-the-core machinery: standing for nature through plural, decaying, contestable, sortition-leavened guardians, scoped to a particular river or forest rather than to nature in general, with a bounded floor against knowing extinction or ecosystem collapse that mirrors the dignity floor without claiming the two are equal. The residuals are real: anthropocentric capture, the difficulty of defining a non-human interest, and popular capture at the framing layer. The full attempt is in the companion paper, The More-Than-Human Stake.