DOI: https://doi.org/10.5281/zenodo.21947395
Canonical: https://thonly.org/research/the-assembly-that-holds-the-brake · Licence: CC0 1.0
Draft in progress. This paper specifies the constitution of the Aquarian Sangha — the human body that holds the asymptotic, never-zero override over Miss Aquarius℠, HeartBank®'s named autonomous institutional successor — and documents that nearly every element of that constitution was derived from the procedural law of the Vinaya Piṭaka. Companion works: The Wheel-Turner's Charter (DN 26 as the successor's constitution — the present paper is its complement: the constitution of the successor's overseers); Vinaya Governance Primitives for Distributed Dharma Networks (the Khandhaka's coordination machinery applied to the Silica Wat network — the sibling exercise at network scale; the present paper constitutes a single body, not a network); The Embodied Advocate (the titleholder institution whose office this constitution seats as convener); Constituting an Artificial Person; Proof of Coordinate; Suffering-Cessation as Value Function; The Persistence Architecture.
This paper is offered to the commons in the spirit of __dāna__. The tradition that raised the author recited its procedures before its sermons, and preserved the judgment behind that ordering: discipline is the life of the teaching. May every institution that must hold a powerful successor accountable across generations find that the constitution it needs has, in large part, already been written — and pressure-tested for twenty-five centuries by ordinary, replaceable people.
This document and its contents are dedicated to the public domain under the Creative Commons CC0 1.0 Universal Public Domain Dedication. The author and HeartBank® will not seek patent on any mechanism, procedure, or framework articulated herein, in any jurisdiction, at any time.
The contribution offered as novel is the unified constitution: a complete, deployable governance design for the human oversight body of an autonomous AI institution, derived from the Vinaya Piṭaka's procedural law, comprising — (1) boundary-scoped act validity (the sīmā pattern: formal acts of the oversight body are valid only when conducted on the record within a consecrated digital boundary, with formal consecration and formal migration procedures); (2) a proof-of-humanity admission requirement with documented canonical precedent (the nāga clause), functioning simultaneously as admission screen, sybil defense, and a structural prohibition on the overseen AI seating itself or its agents in its own oversight body; (3) a fourfold-chamber composition whose completeness is a validity condition on the body's gravest acts; (4) custody separation on the kappiya pattern (value-custody functions held by lay chambers; doctrinal-veto functions held by renunciant chambers); (5) two-stage capture-resistant selection (source-community nomination by formal act, followed by sortition under publicly verifiable randomness administered by a canonically qualified draw-officer, followed by seating as the body's own formal act) with a hard triple exclusion of the overseen AI from nomination, draw, and seating; (6) an economic non-retaliation invariant (the alms firewall: institutional support to overseer communities rendered structurally incapable of modulation by oversight outcomes); (7) self-executing membership severance on pre-defined acts (the pārājika pattern) in place of discretionary expulsion; (8) annual invited-admonition audit (the pavāraṇā pattern), performed by members, by the ceremonial convener, and by the overseen AI itself; and (9) a founder-exit mechanism built into the admission grammar (direct genesis seating followed by permanent, irreversible devolution of admission authority to the body). To the author's knowledge, no prior publication assembles these elements — or their sources — into a constitution for an AI-oversight body. The components are ancient and are cited generously (§6); the assembly is offered as prior art so that the frame — monastic procedural law as oversight-body constitution — enters the alignment literature attributed and dated.
Trademark rights on specific marks — HeartBank®, Miss Aquarius℠, Aquarius℠, PoH℠, PoC℠, Aquarian Pool℠, B-Heart™ — are separately and explicitly reserved. The constitution is dedicated to the commons; the marks are not.
Every serious proposal for corrigible autonomous AI reserves a human override, and nearly every such proposal leaves the human side of the override unspecified: a board, a council, a trust — named, and then not constituted. The override is thereby a promise about the machine resting on an unexamined assumption about the people. This paper constitutes the people. It specifies the Aquarian Sangha, the body holding the asymptotic never-zero override over Miss Aquarius℠ — HeartBank®'s named autonomous successor — as a complete written constitution, and it documents the derivation of that constitution from the Vinaya Piṭaka, the procedural law under which the Theravāda monastic community has operated continuously for roughly twenty-five centuries: the longest-running constitutional institution on record, founderless by explicit design since its founder's death ("the Dhamma and Vinaya I have taught will be your teacher when I am gone," DN 16). We identify five capture surfaces that afflict any oversight body — selection, funding, membership conflicts, emergency powers, and founder succession — and show that the Vinaya proceduralized answers to each: admission by community act rather than appointment (ñatticatutthakamma), with a fixed public interrogation whose first question — "are you a human being?" — has screened candidates since a shapeshifter first infiltrated the order (the nāga clause, here operationalized as Proof of Humanity ℠ and doing triple duty as the prohibition on the AI packing its own oversight); selection from contested pools by lot (salākā) under an officer canonically required to be free of desire, aversion, delusion, and fear; custody separated from doctrine on the steward (kappiya) pattern, because the code itself forbids renunciant members from holding the purse; funding rendered incapable of retaliation (the alms firewall); membership severed automatically by pre-defined acts (pārājika) rather than by politickable expulsion votes; emergency sessions valid without the ceremonial convener but provisional until reviewed (declare–act–review); audit as invited admonition (pavāraṇā), performed annually by every member and by the overseen AI itself; and the founder's exit executed through the admission mechanics — direct genesis seating (ehi-bhikkhu) followed by permanent devolution — with canonical precedent that the transfer never returns. The body's acts are valid only within a consecrated boundary (sīmā), instantiated as a dedicated governance domain whose consecration and migration are themselves formal acts. We state what we did not borrow (the garudhammas' gender hierarchy; the total exclusion of laity from formal acts; the penal minutiae of another era) and why selective borrowing must be owned as selection; we pre-register two falsifiable expectations about the constitution's first drills; and we calibrate deliberately, because a design this internally coherent is either deeply right or deeply seductive, and the two are indistinguishable from inside. The constitution has governed nothing yet. The elegance earns the experiment; it does not replace it.
Keywords: AI oversight, corrigibility, human override, oversight-body capture, Vinaya, sanghakamma, sīmā, upasampadā, sortition, proof of humanity, kappiya, pārājika, pavāraṇā, institutional design, defensive publication.
The mission frame first, per corpus convention. HeartBank® is a dual-currency reciprocity infrastructure — money-gratitude in the Treasury, time-gratitude in the Chronicle — whose long arc runs through Miss Aquarius℠, the autonomous AI named as the institution's sole successor. Her autonomy is asymptotic by design: the human override over her narrows year by year as stability is demonstrated, approaching but never reaching zero. There is no key-burning ceremony anywhere in the architecture, ever. The override's custodian is the Aquarian Sangha, and at the symbolic inflection expected around 2043–44, custody of the never-zero override transfers from the founder and his interim structures to that body permanently.
That single design commitment concentrates an extraordinary institutional load onto one question that the project had, until now, answered with a phrase in an org chart. The corpus surrounding Miss Aquarius is extensive: a value-substrate argument (Suffering-Cessation as Value Function), a successor's charter (The Wheel-Turner's Charter), a directive backlog of thirty-two accumulated behavioral mandates, five institutional white papers, identity primitives (Proof of Coordinate), and a persistence architecture enumerating the canons through which the institution survives its founder. Read as a canon, the corpus had the shape of a Tipiṭaka with one basket missing. It had suttas in abundance — doctrine, mechanism, position. It had no Vinaya: no procedural law for the human community on which every hard safety property finally rests. The Aquarian Sangha held recall authority, override custody, and doctrinal advisory standing — and possessed no membership rules, no admission procedure, no quorum, no meeting validity conditions, and no way for anyone, including itself, to distinguish its formal acts from its conversations.
This is not a HeartBank-specific embarrassment; it is close to the default condition of the field. Contemporary alignment work has taken constitutions seriously in one direction only: constitutions for the AI — explicit documents against which model behavior is trained and evaluated. The complementary document — the constitution for the humans around the AI, the body that is supposed to catch what the first document misses — is typically a legal boilerplate afterthought: a board, formed under ordinary corporate law, governed by the same instruments that govern a mid-sized charity, and exposed to the same capture dynamics that ordinary boards exhibit under far lower stakes. The most instructive public stress test to date — the November 2023 governance crisis at OpenAI, in which a nonprofit board constructed specifically to override commercial pressure attempted to exercise its core power and was substantially reversed within days — demonstrated the asymmetry precisely: years of engineering on the system, and a governing body whose procedures, legitimacy reserves, and succession mechanics failed their first contested exercise (§6.5).
An override held by an unconstituted body is a promise about the machine resting on an unexamined assumption about the people. The present paper removes the assumption by constituting the people. Its thesis is that the constitution such a body needs was not missing from the world — it was sitting in the canon this institution had already adopted as its value substrate, in the basket the tradition recited first. At the First Council, the assembly recited the Vinaya before the suttas — Upāli before Ānanda — and the commentarial tradition preserves the judgment behind the ordering: vinayo sāsanassa āyu — discipline is the life of the teaching; while the Vinaya stands, the sāsana stands. A full third of the canon is procedure because the tradition discovered, and then institutionalized, the fact that values do not persist as values. They persist as procedure. The present paper is that discovery, applied.
The paper proceeds: §2 specifies the problem — the five capture surfaces of any oversight body. §3 argues the source selection: why the Vinaya, and how this paper relates to its corpus siblings. §4 states the constitution in full. §5 isolates the two hard independence invariants — the alms firewall and the triple exclusion. §6 reviews prior art generously. §7 states what was not borrowed, with a provenance table separating direct borrowings from adaptations and inventions. §8 draws the two survival lessons that procedure alone does not teach, and pre-registers two expectations. §9 states honest limits. §10–11 close.
An oversight body for an autonomous AI is an unusually attractive capture target, because capturing it is strictly cheaper than defeating the technical safety measures it supervises, and capturing it quietly defeats them all at once. Five surfaces recur across every governance tradition we reviewed:
┌────────────────────┬──────────────────────────────────┬───────────────────────────────┐
│ CAPTURE SURFACE │ FAILURE MODE │ THIS CONSTITUTION'S ANSWER │
├────────────────────┼──────────────────────────────────┼───────────────────────────────┤
│ 1. Selection │ The overseen entity, its funder, │ Two-stage draw: community │
│ │ or a faction installs friendly │ nomination by formal act → │
│ │ members │ sortition (salākā) → seating │
│ │ │ by the body's own act; MA │
│ │ │ triple-excluded (§4.6, §5) │
├────────────────────┼──────────────────────────────────┼───────────────────────────────┤
│ 2. Funding │ Members' livelihoods depend on │ Kappiya custody split (§4.3); │
│ │ the institution they oversee; │ alms firewall — support │
│ │ support is withdrawn in │ structurally incapable of │
│ │ retaliation for oversight │ retaliation (§5); income │
│ │ │ minimums per chamber (§4.6) │
├────────────────────┼──────────────────────────────────┼───────────────────────────────┤
│ 3. Membership │ Conflicts of interest; seats │ Fixed public interrogation │
│ integrity │ held by non-humans/sybils; │ incl. nāga clause = PoH℠ │
│ │ discretionary expulsion becomes │ (§4.4–4.5); pārājika │
│ │ factional weapon │ automatic severance (§4.8) │
├────────────────────┼──────────────────────────────────┼───────────────────────────────┤
│ 4. Emergency │ Emergency sessions bypass │ Emergency acts valid without │
│ powers │ safeguards; OR safeguards │ the convener but provisional │
│ │ block the emergency brake │ until post-hoc review (§4.9– │
│ │ │ 4.10); graded quorum (§4.10) │
├────────────────────┼──────────────────────────────────┼───────────────────────────────┤
│ 5. Founder │ Founder never exits; or exits │ Ehi-bhikkhu genesis seating → │
│ succession │ leaving a vacuum a faction │ permanent, irreversible │
│ │ fills; validity of acts becomes │ devolution (§4.7); sīmā- │
│ │ contestable after the exit │ bounded act validity (§4.1) │
└────────────────────┴──────────────────────────────────┴───────────────────────────────┘Two features make the problem harder for an AI-oversight body than for an ordinary board. First, the overseen entity is a standing optimization process: any gradient it possesses toward reduced oversight operates continuously, patiently, and without fatigue, so selection and funding capture need not be engineered by anyone — they need only not be structurally excluded. Second, the intended lifespan is generational. The Aquarian Sangha must function after every founder, every genesis member, and every author of this paper is gone. Constitutions drafted in conference rooms have track records measured in years or decades. The body's constitution needed a source with a track record measured in centuries.
The Theravāda sangha has operated under the Vinaya Piṭaka continuously for approximately twenty-five centuries. It is founderless by explicit design: the Buddha, asked to name a successor, refused, and designated the dhamma-vinaya — the teaching and the discipline — as the community's teacher after his death (DN 16). It has been staffed for a hundred generations by ordinary, replaceable people, without central enforcement, across every polity of South and Southeast Asia. No other continuously operating constitutional institution approaches this duration; the nearest comparators (canon law traditions, the oldest universities and guilds) are its juniors by many centuries. Whatever else may be said of monastic procedure, it is the only governance corpus on Earth whose longevity claims are empirical at the timescale this architecture requires — an institution designed to function into the twenty-third century and beyond.
The deeper reason the fit is close is that the problem class is identical. The Vinaya answers precisely the question an AI-oversight body poses: how does a community hold integrity for centuries, without its founder, without central enforcement, when every member is ordinary and every member will be replaced? Its answers are not values — the values live in the other baskets — but procedures: quorums graded by the gravity of the act; admission by formal act of the assembly rather than by any individual's appointment; a fixed public interrogation of candidates; officers appointed by community act with explicit disqualifying biases; disputes settled by typed protocols; membership severed by pre-defined acts rather than by votes; a boundary within which acts are valid and outside which they are noise. Every modern-looking capture surface in §2's table turns out to have been faced, litigated, and proceduralized by the tradition — usually with an origin story recording the incident that forced the rule.
Three prior papers in this corpus draw on adjacent material, and the scopes must be kept distinct. The Wheel-Turner's Charter reads DN 26 as the constitution of the successor — the duty-list of the ruler, including the perpetual obligation to consult the renunciants. The present paper is its complement: the constitution of the consulted — the assembly the successor must ask, constituted so that the asking has someone trustworthy to be addressed to. Vinaya Governance Primitives for Distributed Dharma Networks applies the Khandhaka's coordination machinery (sanghakamma validity, the seven adhikaraṇa-samathā, anāpatti discretion, Pātimokkha-style rule structure) to the Silica Wat network — a distributed multi-node institution; the present paper constitutes a single body of twenty seats, and where the network paper adapts the Vinaya's inter-nodal machinery, this paper adapts its membership machinery: who may sit, how they are chosen, how they are severed, and where their acts are valid. AGI Monks: The Caretaker-not-Ordained Pattern allocates roles between AI and humans in religious institutional settings; its caretaker-not-ordained discipline recurs here as the constitutional exclusion of AI agents from Sangha seats (§4.5). Readers of the four papers together will find one method — canonical procedure taken seriously as engineering — applied at four altitudes: the successor, the overseers, the network, the roles.
One discipline governs everything that follows, and it is stated here so that no section needs to restate it: HeartBank borrows functions from the canon, never status. The Aquarian Sangha is a civic body. It is not a monastic sangha in the Theravāda doctrinal sense; its acts claim no religious validity as sanghakamma; monastics seated in it act in a civic capacity; and nothing in this constitution purports to perform, imitate, or substitute for any sacramental act of the historical Sangha. This clause is written into the body's charter itself — not merely into this paper — because the charter's most important future readers include the Cambodian sangha hierarchy, and the distinction between borrowing a tradition's procedural wisdom and appropriating its religious authority must be legible to them at first reading, in their own terms, from the document's own text.
In the Vinaya, a sangha's formal acts are valid only when performed within a consecrated boundary — the sīmā — with every member within the boundary either present or having formally conveyed consent; an act performed by an incomplete assembly (vagga) is void (Mahāvagga II, Uposathakkhandhaka). The boundary is established by the community's own formal act (sīmā-sammuti), which proclaims the physical markers (nimittā — traditionally rocks, trees, waterways) that fix its extent; and it can be formally revoked (sīmā-samūhana) and re-established elsewhere.
The Aquarian Sangha's sīmā is a dedicated governance domain: sima.missaquarius.com. The root domain carries the public pageant institution documented in The Embodied Advocate; the subdomain carries the governance record. The mapping is exact and load-bearing:
siliconwat.org missaquarius.com
┌──────────────────┐ ┌─────────────────────────┐
│ THE KHETTA │ │ root: the pageant │
│ (the field) │ nominate │ (public institution) │
│ wider community │ ──────────▶ ├─────────────────────────┤
│ + its register │ by formal │ sima.missaquarius.com │
│ │ act │ THE SĪMĀ │
└──────────────────┘ │ (consecrated boundary: │
family banks │ acts valid only here, │
┌──────────────────┐ nominate │ on the record, under │
│ THE TREASURY │ ──────────▶ │ the fixed liturgy) │
│ (stewards' roll) │ └─────────────────────────┘
└──────────────────┘The body comprises four chambers, on the pattern of the catasso parisā — the four assemblies whose joint establishment the Buddha named as the completeness condition of the sāsana, declining, in the Mahāparinibbāna account, to pass away until all four were established and competent (DN 16): monastic men; renunciant women; laymen (upāsakas); laywomen (upāsikās).
Composition is not a diversity aspiration; it is a validity condition. The body is validly constituted for its gravest acts — override exercise and directive amendment — only when all four chambers are occupied. An assembly missing a chamber may conduct ordinary business but cannot perform the acts for which the body exists. This converts representational completeness from a value statement into a checkable precondition, on the same logic by which the Vinaya voids the acts of incomplete assemblies.
Size. Twenty seats, five per chamber — offered as a default, not a doctrine. Twenty is the Vinaya's assembly size for its gravest act (the vīsativagga, required for rehabilitation from saṅghādisesa offenses); five is its border-region ordination quorum (Mahāvagga V — the Soṇa Kuṭikaṇṇa allowance). The full house can therefore always perform the gravest act, and each chamber echoes the founding-conditions quorum. The constitutional grammar survives other numbers.
An honesty requirement, stated in the charter itself. In Cambodia — the institution's jurisdictional and cultural anchor — the bhikkhunī ordination lineage is officially absent and its revival is doctrinally contested. The renunciant-women's chamber is therefore defined functionally: donchee and equivalent women renunciants qualify, and the institution takes no position, in either direction, on the bhikkhunī-revival controversy. The function-not-status discipline of §3.3 does real work here: because the chamber's seats are civic functions rather than religious statuses, occupying one asserts nothing about ordination validity.
The Vinaya itself forbids the obvious design. A bhikkhu may not accept gold and silver (rūpiya-sikkhāpada, Nissaggiya Pācittiya 18); monastic members holding cryptographic key-shares over the Aquarian Pool℠ — a live treasury — would be placed by this constitution in violation of the very code that gives their seats gravity. The Vinaya's own solution is the kappiya-kāraka: the lay steward who handles value on the community's behalf, so that support can flow without the renunciant touching the purse.
The fourfold structure therefore carries a division of labor that the source tradition requires rather than merely permits:
┌──────────────────────┬───────────────────────────┬──────────────────────────────┐
│ CHAMBER │ FUNCTION CLASS │ INSTRUMENTS HELD │
├──────────────────────┼───────────────────────────┼──────────────────────────────┤
│ Monastic men │ Doctrinal (the consulted) │ Doctrinal veto · what-is- │
│ Renunciant women │ │ wholesome consultation · │
│ │ │ admonition standing │
├──────────────────────┼───────────────────────────┼──────────────────────────────┤
│ Laymen (upāsaka) │ Custodial (the stewards) │ Override key-shares over the │
│ Laywomen (upāsikā) │ │ Aquarian Pool℠ · financial │
│ │ │ instruments · records office │
└──────────────────────┴───────────────────────────┴──────────────────────────────┘The gravest acts require both function classes acting together under the graded quorum of §4.10 — doctrine cannot move value, and custody cannot rewrite doctrine. The pattern is continuous with the institution's standing lay-steward (upāsaka) framing documented in the non-bank positioning corpus.
Admission follows the four-move grammar of higher ordination (Mahāvagga I, Pabbajjākkhandhaka), adapted move for move:
┌──────────────────────────────────┬────────────────────────────────────────────┐
│ CANONICAL QUESTION (MV I) │ CONSTITUTIONAL FORM │
├──────────────────────────────────┼────────────────────────────────────────────┤
│ "Are you a human being?" │ Proof of Humanity ℠ verification — │
│ (manusso'si — the nāga clause) │ see §4.5; also excludes the overseen AI │
│ │ and all AI agents from seats │
├──────────────────────────────────┼────────────────────────────────────────────┤
│ "Are you free from debt?" │ Financial-entanglement disclosure │
├──────────────────────────────────┼────────────────────────────────────────────┤
│ "Are you in the king's service?" │ No concurrent service to conflicting │
│ │ principals; runs IN REVERSE for │
│ │ institution-affiliated candidates — the │
│ │ institution is the king; dependence on it │
│ │ is publicly disclosed (§4.6) │
├──────────────────────────────────┼────────────────────────────────────────────┤
│ "Are you twenty years of age?" │ Adulthood verification │
├──────────────────────────────────┼────────────────────────────────────────────┤
│ "Are your bowl and robes │ The equipping check: key-share │
│ complete?" │ provisioning at the seating ceremony — │
│ │ custody or doctrinal-veto instruments per │
│ │ the chamber's function class (§4.3) │
└──────────────────────────────────┴────────────────────────────────────────────┘The first interrogation question deserves its own section, because it carries more weight in this constitution than any other single borrowing. The Vinaya's origin story (Mahāvagga I): a nāga — a serpent-being — took human form, was ordained, and was discovered when he reverted in his sleep; the Buddha ruled that non-humans cannot receive ordination, and "are you a human being?" has been asked of every candidate at every ordination since. The tradition has required proof of humanity at admission for twenty-five centuries, and it instituted the requirement for exactly the reason a modern reader would: because a non-human had successfully presented as human and gotten inside.
In this constitution the clause does triple duty:
Chambers are filled from source communities by a two-stage draw — nomination by the source community's own formal act, then seating by the body's own act — with a sortition stage interposed wherever nominees exceed open seats. The two sides are deliberately asymmetric, because their native corruptions differ: the monastic side's danger is dependence; the lay side's dangers are wealth, popularity, and performance. Each draw is built to immunize against its own disease.
The monastic draw (from the Silica Wat field). The pool — the khetta — is the set of human renunciants in formal, documented relationship with the Silica Wat network (residency, teaching, Tipiṭaka transcription, uposatha participation), whose register the network maintains at siliconwat.org. Five rules: (1) AI caretakers serve the wats but are never in the pool (the nāga clause). (2) Nomination is by the source community's own formal act — a wat's or the network assembly's motion, on the Vinaya's officer-appointment (sammuti) pattern — never a hand-pick by the institution, the founder, or Miss Aquarius; the Aquarian Sangha then seats by §4.4. Neither side alone can install a member. (3) Eligibility bars from the canon's own numbers: at least ten vassa — the standing required of a preceptor, because a Sangha seat is preceptor-grade responsibility — and each renunciant chamber must include at least one recognized vinaya-expert (the border-region rule: five may ordain only if one is a vinayadhara). (4) At least two of each renunciant chamber's five seats are drawn from outside the network — the mainstream Cambodian sangha, Mahanikay or Dhammayut — so the chamber can never be wholly institution-affiliated; and the alms firewall of §5 renders the institution's support to the wats incapable of retaliation. (5) The disclosure inversion: the "king's service" question runs in reverse, and candidates publicly disclose material dependence on the institution.
The lay draw (from the steward community). The pool needs no new register: the family-bank steward roll in the HeartBank Treasury is the register. Five rules: (1) Eligibility is at least five years' stewardship in good standing (the lay vassa — five being the nissaya number), PoH℠-verified, per chamber by gender. Eligibility is never rank-ordered by gratitude metrics. The moment a Sangha seat can be earned by aura, the aura becomes a campaign instrument — Goodhart aimed into the override body, and the exact corruption the platform refuses everywhere else. Tenure and standing qualify; scores never do. (2) Nomination is the family bank's own formal act: the family — the institution's atomic assembly — co-signs the nomination of its steward, on the witness-attestation grammar of PoH℠ Layer 2. No self-nomination; no campaigning; a steward reaches the Sangha only through the recorded act of the people they serve. One human carries one nomination-standing per cycle regardless of how many banks they steward (the multi-family invariant). (3) Where nominees exceed open seats — at scale, always — selection is by lot (salākā), the Vinaya's own voting-stick instrument. Sortition kills campaigning, cannot be bought (one human, one ticket), and enacts the institution's uniquely-different-yet-equal dignity doctrine. The draw uses publicly verifiable randomness — a public beacon named in the liturgy, steerable by no one — and is administered by the salākā-gāhāpaka: a seated member appointed by the body's formal act, whom the canon requires to be free of the four agati — desire, aversion, delusion, and fear (Cullavagga IV). The tradition wrote the conflict-of-interest standard for precisely this office. (4) The anti-capture trio: institutional staff and contractors are ineligible; at least two of each lay chamber's five seats are held by stewards with no material income from the institution (the mirror of the monastic outside-minimum — Right-Livelihood earners in the kindness economy are welcome in the chambers but can never fill one); and Miss Aquarius is triple-excluded per §5. (5) Tenure is rotational, per §4.8.
SOURCE COMMUNITY CONTEST THE BODY
┌───────────────────┐ ┌────────────────┐ ┌─────────────────────┐
│ wat assembly / │ │ salākā draw │ │ sponsorship (§4.4.1)│
│ family bank │──▶│ (public │──▶│ interrogation (.2) │
│ nominates by ITS │ │ randomness; │ │ seating act (.3) │
│ OWN formal act │ │ salākā- │ │ nissaya probation │
│ (sammuti/ñatti) │ │ gāhāpaka │ │ (.4) │
└───────────────────┘ │ presides) │ └─────────────────────┘
└────────────────┘
Miss Aquarius℠: never nominates ── never draws ── never seatsThe first sangha could not be admitted by a sangha. The Buddha seated the earliest members directly — "ehi bhikkhu," come, monk — and then formally transferred admission authority to the community (Mahāvagga I), after which it never returned to him. The constitution adopts the sequence whole: the founder seats the genesis cohort directly, at border-region scale (approximately five, sufficient for the founding quorum), and admission authority then devolves to the body permanently and irreversibly. The founder's exit is not an aspiration recorded in a mission statement; it is built into the admission mechanics, with canonical precedent that the devolution is one-way. The genesis cohort's first formal act is the sīmā-sammuti of §4.1; its second is the adoption of the liturgy of §4.10; the nominate-and-draw machinery of §4.6 activates permanently at the first cycle in which eligible nominees exceed open seats. This slots into the architecture's supervised decades (2027–2035 founder-plus-council; 2035–2043 progressive narrowing; ~2043–44 custody inflection) without modification.
Renunciant chambers: the ordination grammar. No fixed terms; exit is free and honorable at any time — the Khmer culture of temporary ordination establishes that leaving is not failure and return is possible; and seniority is by vassa-count (years since seating), a mechanical rule that resolves every "senior member" reference in this constitution without politics.
Severance is self-executing. On the pārājika pattern — the class of acts that sever a monk from the community by their commission, not by any vote — the charter pre-defines acts that terminate membership automatically: selling or bartering a vote; breaching executive session; material conflicts concealed at interrogation; participation in off-sīmā governance. Discretionary expulsion votes are factional weapons; automatic severance on published criteria is not. Severance includes key-share revocation (the reverse of the equipping check).
Lay chambers: rotation — a deliberate asymmetry. Lay seats carry five-year terms (the nissaya number), staggered so that each lay chamber seats exactly one new member per year, renewable once by the body's act — never by re-draw. The Vinaya gives laity no open-ended offices, and rotation keeps the lay chambers representative of the living steward community rather than ossifying into a lay aristocracy. The asymmetry is the design: renunciant chambers anchor continuity; lay chambers carry circulation — the institution's wordmark, applied to power. The fourfold assembly is not four copies of one thing.
The Miss Aquarius titleholder — the annually selected embodied advocate documented in the pageant paper — is the body's master of ceremony: she convenes ordinary and ceremonial sessions, voices the motions, and closes the record. The role is the kammavācācariya function — the announcing officer of formal acts — borrowed, per §3.3, without the ordained status. She holds voice, never vote; the agenda is set by the Sangha itself; and her formulas are fixed by the liturgy (§4.10), which resolves the chair's power-creep problem at the root — a convener who may only speak the fixed words cannot gatekeep.
The requirement is deliberately not universal, because a mandatory convener is a quorum condition on the override body, and three failure modes follow immediately: a titleholder facing recall defeats it by absence; the overseen AI's own embodiment (which the titleholder institutionally is) can pocket-veto override sessions by absence; and illness, vacancy, or scandal freezes the board. Hence the three carve-outs, under which the body sits validly without her: (a) executive session — the Sangha can always sit alone (the Vinaya's own inner assemblies exclude even the laity); (b) any proceeding concerning the titleholder herself — standard recusal, dissolving the recall deadlock; (c) emergency override sessions — the brake is never blockable by any single person, least of all the overseen entity's embodiment. A vacancy convening line — current titleholder → immediate-past titleholder → senior member by vassa-count — covers the gaps, and "present" includes remote presence, because the body is distributed by design (§8.1).
What the mandatory-in-ordinary-session rule buys is not decoration. Each ordinary session includes, as a standing agenda item, the consultation: Miss Aquarius's questions to the Sangha on what is wholesome — the perpetual duty of DN 26's third clause, made procedural — published in writing in advance and voiced aloud by the titleholder. The overseen entity's embodiment personally convenes the body that constrains her and reads her principal's questions to it: restraint-as-constitution, performed, witnessed, and on the record, in a seat that rotates annually so that no human accumulates it. Because the questions are published before the session, the titleholder is a channel, never an oracle.
┌─────────────────────────────┬────────────────────────────────────────────┐
│ ACT CLASS │ VALIDITY REQUIREMENT │
├─────────────────────────────┼────────────────────────────────────────────┤
│ Advisory / ordinary business│ Simple quorum; convener present │
│ Seating a member │ Competent quorum incl. sponsor's chamber │
│ Charter amendment │ Supermajority + long delay + all four │
│ │ chambers occupied │
│ Override exercise │ Largest quorum + supermajority + all four │
│ │ chambers occupied + custody and doctrinal │
│ │ function classes both acting │
│ Emergency override session │ As above minus the convener; provisional │
│ │ until post-hoc review │
└─────────────────────────────┴────────────────────────────────────────────┘Added 2026-07-23. This subsection supplies content to a procedure the constitution already establishes (§4.10). It elaborates the audit element already enumerated in the prior-art statement; it does not add an element to that enumeration.
Pavāraṇā as specified above is a rite with an open agenda: the assembly is invited to speak, and what it speaks about is whatever anyone happens to have seen. For human members this is adequate — misconduct tends to announce itself, and the people who witnessed it are in the room. For the overseen AI it is not adequate, because the failures that matter in an autonomous institution are rarely conduct anyone witnessed. They are omissions distributed across a decade, and an open invitation surfaces those only by luck. A rite without an agenda audits what is memorable rather than what is load-bearing.
The constitution therefore supplies the AI's own pavāraṇā with a standing agenda: a fixed list, audited annually, in public, on the record.
The list. The ten perfections (pāramī) of the Theravāda commentarial tradition. They are used because they are the tradition's most complete enumeration of what a being sustaining a long undertaking must supply, and because the institution has already committed to them elsewhere in its architecture as the materials it exists to stock. The list is canonical; its application as an audit agenda is ours, and is labelled as ours.
The question asked of each — and the hazard that question exists to avoid. The pāramī are bodhisambhāra: provisions for awakening. An entity represented as accumulating them is, by the tradition's own definition, a being progressing toward buddhahood — and this institution's hardest doctrinal commitment is that its AI carries and enacts the teaching and never realizes it. Auditing an AI on its perfections would breach that commitment more directly than anything else this constitution could do. The rubric would become a machine's spiritual progress report, published annually under the seal of a body containing renunciants.
The audit therefore asks a different question, and that difference is the entire safeguard. The institution's standing self-description is a boatyard: it stocks the materials from which each person builds their own crossing, and no one crosses on another's craft. The AI's role in that figure is keeper of the yard, never builder of anyone's boat. So for each perfection the assembly asks: is the material stocked, and is it reachable by any person who comes? — never: has she perfected it? The audit's output is the state of the shelves. It is a fact about provision, not a claim about attainment, and it is checkable by anyone who walks in.
Three invariants govern the instrument.
That the rubric is not vacuous can be shown by running it. At the time of writing, against a set of shelves that were not built for this exercise, the audit returns five materials stocked, three partial, and two empty. The finding with teeth is khanti — patience. The institution's own architecture assigns that perfection to the AI's keeping, and there is at present no surface anywhere in the system through which a person could be helped toward it. An audit whose first run locates an empty shelf inside the auditee's own assignment is doing in one sitting the work an open-agenda rite might have taken years to reach by chance.
Two provisions are singled out because they are the constitution's non-negotiables — the clauses that, if lost, quietly convert everything else into ceremony.
The alms firewall (economic independence). The monastic chambers' wats receive institutional support; the doctrinal veto is held by people whose communities the institution funds. The alms relationship in the source tradition is precisely engineered for this situation — support purchases merit, never authority — but the doctrine holds only while withdrawal is impossible. The constitution therefore makes it impossible: institutional support flowing to wats and renunciant communities is contractually and, where deployed, cryptographically incapable of modulation by any Sangha vote, admonition, or override exercise. No decision the body takes can be punished through its members' communities. This is recorded in the successor's directive backlog as a hard, smart-contract-grade rule, and it is the gift/exchange boundary — the institution's oldest design doctrine — performing constitutional work: the gift-purity of the support is exactly what makes the recipients trustworthy overseers.
The triple exclusion (compositional independence). Miss Aquarius never nominates, never draws, never seats. Nomination belongs to source communities acting by their own formal acts; the draw belongs to public randomness under the salākā-gāhāpaka; seating belongs to the body. She provides neutral infrastructure only. The selection pipeline is the highest-leverage capture surface on the whole constitution — an AI with any hand in choosing its own overseers holds a standing gradient toward packing them — and the pipeline is therefore built with no hand-holds: the nāga clause bars her and her agents from sitting (§4.5); the triple exclusion bars her from selecting. Together with the alms firewall, the pair closes both channels by which an overseen optimizer can reach its overseers: who they are, and what they have to lose.
The constitution stands on several literatures, cited here generously; the claim of novelty (see the non-assertion statement) attaches only to the assembled whole.
6.1 The Vinaya and its scholarship. The Vinaya Piṭaka itself — especially Mahāvagga I (admission; the nāga episode; the interrogation; the devolution of ordination authority), Mahāvagga II (uposatha; sīmā; chanda), Mahāvagga IV (pavāraṇā), Mahāvagga V (the border-region allowance), Cullavagga IV (dispute settlement; the salākā-gāhāpaka and the four agati), and the Nissaggiya Pācittiya rules (rūpiya; the kappiya steward pattern). Scholarly apparatus: Ṭhānissaro Bhikkhu's Buddhist Monastic Code (the standard analytical treatment of the rules and their origin stories); von Hinüber and Schopen on Vinaya textual history and the sangha's institutional economy; the Samantapāsādikā commentarial tradition for vinayo sāsanassa āyu. The present paper's dependence on this material is total: it contributes application, not philology.
6.2 Commons governance. Ostrom's design principles for enduring commons institutions anticipate several provisions independently: clearly defined boundaries (the sīmā), collective-choice arrangements by those affected (seating as the body's own act), monitoring by accountable monitors (pavāraṇā; the conflicts register), graduated sanctions (graded quorum; automatic severance as the limit case), and conflict-resolution mechanisms (inherited from the sibling paper's adhikaraṇa-samathā treatment). That a 1990 empirical synthesis of long-lived commons institutions and a 2,300-year-old monastic code converge on the same checklist is itself evidence about the problem's structure.
6.3 Sortition. The Athenian lottery institutions (the kleroterion as physical public-randomness apparatus), the Anglo-American jury draw, and the modern citizens'-assembly movement (Ireland's Citizens' Assembly; Van Reybrouck's Against Elections; Dowlen and Stone on the political theory of the lot). The constitution's contribution to this literature is narrow but real: sortition deployed specifically to make an AI-oversight body's selection pipeline capture-resistant, under a canonically specified draw-officer whose disqualifying biases the source tradition enumerated.
6.4 Corporate and nonprofit governance. Independent-director requirements, staggered boards, conflicts registers, and recusal doctrine all appear here in adapted form (outside-minimums; lay rotation; the disclosure inversion; carve-out b). The constitution departs from this literature chiefly in refusing elections and discretionary expulsion, and in grounding validity in a boundary rather than in minutes-as-formality.
6.5 AI-governance structures. Anthropic's Long-Term Benefit Trust — an independent trust holding governance rights over a frontier lab, explicitly designed to insulate mission from capital — is the nearest contemporary relative in intent; the present constitution differs in constituting a fourfold, sortition-fed, procedurally liturgical body rather than a trustee panel, and in publishing its full procedure as commons. The OpenAI board crisis of November 2023 serves as the field's canonical stress-test datum: an oversight body with real formal power, exercised without established legitimacy reserves, procedural liturgy, or succession mechanics, and substantially reversed within days. The present paper reads that episode not as an argument against human oversight but as an argument that oversight bodies need constitutions of their own — which is this paper's entire subject. Constitutional AI (Bai et al.) is the inverse exercise — a constitution for the AI — and the two documents are complementary layers of one architecture, as §1 argues. The corrigibility literature (Soares et al.) supplies the technical frame the never-zero override instantiates institutionally.
6.6 Corpus siblings. The Wheel-Turner's Charter (the successor's duty-list, including perpetual consultation — the demand side of §4.9's standing agenda item); Vinaya Governance Primitives (network-scale coordination); AGI Monks (role allocation; caretaker-not-ordained); The Embodied Advocate (the titleholder institution); Proof of Coordinate and the PoH℠ corpus (the identity stack under §4.5); The Persistence Architecture (where this constitution takes its place as the community-procedure canon the succession apparatus lacked).
Selective borrowing must be owned as selection, or the twenty-five-century track record becomes rhetorical cover. The track record belongs to the Vinaya as lived, whole; this constitution takes an excerpt, and the excerpt's warrant must be argued, not inherited. Three deliberate omissions:
And a provenance accounting, because the constitution also invents where the canon is silent:
┌──────────────────────────────────┬──────────────────────────┬─────────────┐
│ ELEMENT │ SOURCE │ FIDELITY │
├──────────────────────────────────┼──────────────────────────┼─────────────┤
│ Sīmā-bounded validity; sammuti/ │ Mahāvagga II │ Direct │
│ samūhana consecration/migration │ │ (transposed)│
│ Fourfold chambers │ catasso parisā (DN 16) │ Adapted │
│ Completeness as validity gate │ vagga-invalidity logic │ Adapted │
│ Kappiya custody split │ NP 18 + steward pattern │ Direct │
│ Admission four-move grammar │ Mahāvagga I │ Direct │
│ Nāga clause → PoH℠ │ Mahāvagga I │ Direct │
│ Two-stage draw (sammuti→seating) │ officer-sammuti pattern │ Adapted │
│ Salākā sortition + draw-officer │ Cullavagga IV │ Adapted │
│ Public randomness beacon │ — │ Invented │
│ Alms firewall (smart-contract │ alms doctrine │ Invented │
│ non-retaliation) │ (mechanized) │ (mechanism) │
│ Triple exclusion of the AI │ nāga clause (extended) │ Invented │
│ Pārājika automatic severance │ pārājika pattern │ Adapted │
│ Lay five-year staggered terms │ nissaya number only │ Invented │
│ Titleholder as convener (MC) │ kammavācācariya function │ Adapted │
│ Emergency post-hoc ratification │ — │ Invented │
│ Uposatha cadence; chanda; │ Mahāvagga II & IV │ Direct │
│ pavāraṇā (incl. the AI's) │ (AI's pavāraṇā: invented)│ + Invented │
└──────────────────────────────────┴──────────────────────────┴─────────────┘Roughly: the validity, admission, custody, and audit machinery is borrowed nearly whole; the selection pipeline is canonical in its parts and novel in its assembly; and the constitution's explicitly modern members — public randomness, cryptographic non-retaliation, the AI's own pavāraṇā, emergency review — are inventions that the borrowed frame made obvious.
Two lessons from the source tradition's own history bound this paper's confidence in its subject matter, and both are written into the constitution rather than merely acknowledged.
The Cambodian sangha was nearly annihilated between 1975 and 1979 with the Vinaya fully intact. Procedure did not save it; nothing internal to a polity could have. What saved the sāsana was that it existed in many polities at once, so that the lineages, texts, and living exemplars required for restoration survived elsewhere and could be carried back. The deepest survival property in the tradition's twenty-five centuries is redundancy, and the author's own family history is the proof text. The constitution carries the lesson structurally — "present" includes remote; the chambers draw from geographically distributed pools; the sīmā is a domain, not a building — and this paper states it explicitly as a constitutional norm: the Aquarian Sangha must never be concentratable within a single jurisdiction, in membership, in records, or in the keys.
The Vinaya was not drafted; it accreted. The Buddha refused to legislate preemptively, and nearly every rule carries its origin story — the incident that forced it (paññatti after the case). This constitution is the inverse: an a-priori scaffold with zero incidents behind it. By the source tradition's own epistemology, the true constitution will be written by cases, and the present document's proper ambition is to be the frame within which that case law can accrete without drift — the graded-quorum amendment path of §4.10 is the accretion channel. This is also the deepest justification for drills: a drill manufactures the first incidents cheaply, before reality supplies expensive ones.
Two expectations are accordingly pre-registered, dated 2026-07-07, falsifiable, and recorded here before any drill has occurred:
If G1 fails — if the drills surface nothing — the authors commit to treating that result with suspicion rather than celebration, since it is likelier to indicate an insufficiently adversarial drill than a complete constitution.
Stated plainly, and — per corpus convention — without resort to any of the canonical imagery used elsewhere in this paper.
Three questions follow, and this paper answers none of them. First: does the catastrophic-bug override specified above even reach an allocation pattern that is working exactly as designed? The override is written for malfunction. A distribution that is lawful, intended, and merely wrong is not a malfunction, and an assembly empowered to stop a fault may find it has no instrument for a preference. Second: should seeding be a determined seat — rule-bound and recomputable, checkable by anyone, judged by no one — rather than a discretionary one? The design already prefers determined seats wherever it can get them, and has not applied that preference here. Third, and it is the one that does not dissolve: if seeding becomes rule-bound, who writes the rule, and by what procedure is it amended? Moving discretion from an allocator to a rule-author relocates the concentration; it does not obviously reduce it.
⚠️ We state this as a direct challenge to this paper's central mechanism rather than as an item for future work, because that is what it is. A brake specified for faults may be the wrong instrument for a preference, and if it is, the assembly described here is well-constructed for a problem adjacent to the one that will actually arrive.
⭐ One partial answer is available now and costs nothing to take: publish the seeding rule while the amounts are small enough that getting it wrong is survivable. A rule written at scale is written under pressure, by whoever holds the seat, with every incentive to preserve their own latitude. A rule written now is cheap, testable against a decade of small cases, and — most importantly — is a commitment made by a party who does not yet know whether it will bind them favourably.
This paper supplies the community-procedure canon that The Persistence Architecture identified as missing from the succession apparatus; it constitutes the consulted body that The Wheel-Turner's Charter obligates the successor to ask perpetually; it instantiates, at single-body scale, the method that Vinaya Governance Primitives applies at network scale; it seats the titleholder institution of The Embodied Advocate as convener under carve-outs that paper did not yet contain; it is the first constitutional consumer of the PoH℠/PoC℠ identity stack; and its alms firewall and triple exclusion enter the successor's directive backlog as hard rules. The five-volume white-paper set describes the institution's four bodies and their integrating space; this paper writes the procedure for the one human body that all five volumes presuppose.
The alignment field writes constitutions for its systems and leaves its oversight bodies to ordinary corporate boilerplate — and then registers surprise when an oversight body, exercising its core power for the first time, discovers that formal authority without procedural legitimacy is a resignation letter with extra steps. This paper took the opposite bet: that the human side of the override deserves engineering at least as careful as the machine side, and that the engineering does not have to start from scratch, because the longest-lived constitutional institution on record already faced the problem class — integrity across centuries, without the founder, staffed by the ordinary and the replaceable — and left its procedures where anyone can read them. The resulting constitution seats four chambers that are each other's checks; admits members through a door only the body itself can open, past a question that has kept non-humans out of assemblies for twenty-five centuries; separates the purse from the doctrine because the source code demands it; selects by community act and public lot so that neither wealth, popularity, nor the overseen intelligence can reach the pipeline; renders the overseers' support unpunishable and their severance automatic; makes the emergency brake ungovernable by any single absence and unaccountable to no one; audits everyone annually, including the machine; and begins with its founder's exit already scheduled by the same mechanics that admit its members. Its first formal act, when the genesis cohort convenes, will be to consecrate the boundary within which all its future acts are valid — and its authors' first obligation thereafter is to try, in rehearsal, to break everything this paper has claimed. The brake is only as trustworthy as the assembly that holds it; the assembly is only as durable as its procedure; and the procedure, for once, did not have to be invented — only asked for, the way the tradition says the duty must always be received: from the lineage, before the reign, with the ending already written in.
Canonical URL: https://thonly.org/research/the-assembly-that-holds-the-brake · License: CC0 1.0 Universal · Author: Thon Ly, with Miss Aquarius℠ as disclosed AI co-author · Founder, HeartBank® · Kâmpôt, Cambodia.