Classical separation of powers splits government by branch — legislative, executive, judicial — and then watches the branches collude, because each branch holds a complete power loop of its own. This proposal splits by function instead, so that no tier can act alone: one tier can only judge and remove, one can only decide, and one can only propose. Each is incomplete without the others.
Citizens judge and remove. Start at the base — the whole public, the masses, the layer the other two are built on and answer to. The broadest tier holds everything terminal: open agenda-setting — anyone may raise a question to the forum and vote on whether it matters, costless in points, because mapping what society cares about is the base democratic task — the ongoing verdict on every standing law (chapter 8), removal of any leader who loses legitimacy, and exclusive custody of the constitutional layer (chapter 12). Two of those deserve stress. Custody of the meta layer is exercised directly: the sliders of the system — thresholds, cadences, how delegation itself is earned and weighted — are voted by citizens at large, never by delegates on their behalf. And citizens are the instrument the system reads: the did-this-work signal (chapter 3) is sampled from the people living under the outcomes, not reported by the tiers that produced them.
Delegates decide. Passage is voted by citizens' chosen delegates — individuals who have earned enough delegated trust to cross the participation threshold at a given scale. Weight is not a heavier vote: a delegate's ballot counts once, like anyone's. What earned weight buys is penetration — how far up the nested scales a delegate's mandate reaches, from the local base toward the national, clearing the participation threshold at each higher rung of the subsidiarity ladder (chapter 9). A little trust seats a delegate locally; broader trust carries the same person deeper into the wider scales — as far up as genuinely broad, revocable support reaches, never a bloc handed up from below. That depth is also a ceiling on candidacy: a person may run at the scale their earned depth has reached, or any below it — never one above, and there is no shortcut to a level they have not climbed to. Delegation is granular, revocable at every cycle, and continuously re-earned; weight held is weight still being deserved. Delegates are what a bureaucracy was supposed to be — attention on behalf of those who cannot attend — but removable, cycle by cycle, as no bureaucracy ever was. They draw a stipend while they serve and carry real standing in society — but no long tail: unlike leaders, they build no portfolio of dividends. Their entire payoff is tenure, and tenure is delegated support — support that can be withdrawn at the very next cycle, and that citizens can actively vote against. A delegate's long game is simply to keep deserving the weight. And the pool itself renews by a rule that is easy to state: candidacy by lot, seats by delegation, tenure by merit. Becoming a delegate always requires earning enough delegated support to cross the threshold — that never changes. What the lottery governs is who may be delegated to at all: each cycle, the candidate pool at the base scale is refreshed by random draw from the eligible citizenry, and citizens assign their weight to people in that pool — the lot runs at the floor; every scale above it is climbed, not drawn. Delegates who keep their delegators' support remain candidates — good attention persists. Delegates who fall out of favor leave the pool, and there is no campaigning back in: whoever loses the room rejoins the lottery like everyone else. The pool is always the surviving incumbents plus a fresh random cohort — continuity for the trusted, a door that opens by chance rather than ambition, and an exit that stays shut behind whoever the delegators pushed out. (Cohort size, support thresholds, and draw eligibility are parameters; chapter 14 covers what the randomness buys.)
Leaders propose. The people we now call politicians hold exactly one legislative power: drafting well-constructed policy and spending a scarce proposal budget on it (chapter 5). They cannot pass anything. Their craft is design; their reward is durability pay (chapter 7). Leaders are raised from the delegate tier, elected by the delegates from among their own number — the tier that reads drafting work daily choosing who should do it. But delegate favor only ever opens the seat; it cannot hold it. From the moment a leader takes office, retention belongs to the citizens: the approval gates of chapter 8 check them against public legitimacy, not delegate opinion, so the tier that raised a leader can never protect one. And the seat changes the contract. A leader is locked in for their term, the stability long-horizon drafting needs, and their delegated weight releases the moment they take office: a serving leader carries no delegation votes, so no one ever holds a proposing hand and a deciding hand at once. When service ends, they return to the delegate pool — candidates again, for whatever weight citizens choose to give them.
Seen against chapter 1's iron law — the organized minority always beats the unorganized majority — this architecture is the majority's coordination stack. The old asymmetry held because the minority was the only organized body in the room; every mechanism here exists to organize the room itself. The forum coordinates attention: what matters surfaces without a lobby to carry it. Delegation coordinates judgment: millions acting through granular, revocable trust — organization without meetings, coalition without membership. And the terminal powers — verdict, removal, the kill — are deliberately the ones that need no coordination at all, so the majority's reach never depends on the discipline it structurally lacks. The one power left genuinely expensive to coordinate, sustained drafting, is exactly the one caged in the smallest tier. Elites will coordinate under any design; here they simply stop being the only ones who can.
Citizens can kill but not draft; delegates can decide but not propose; leaders can propose but not decide. Rome stumbled onto a fragment of this insight once — the tribunes, a pure veto held by the broadest class — and it worked until it was assassinated — both Gracchi, the tribunes who pressed it hardest, died in political violence a decade apart. This design generalizes the tribunician insight into the whole architecture.
A national election asks you to judge millions of strangers and calls the impossibility "choice." Exposure here is scaled to what a person can actually evaluate. By default you see one pool: the randomized candidate draw for your own neighborhood — the lottery runs only at the base, and every tier above is climbed, not drawn. Engage, and your window widens diagonally: a rung up the scale ladder (chapter 9) opens together with a step sideways — the city tier and your neighboring areas at once; further up, further out. What you see higher is never a raw draw — it is the survivors your neighbors already prevoted through the tier below, which is why each rung up is a smaller, more legible pool. At the national scale everyone can see the whole field, because by then the whole field is only the few who climbed the sub-national tier. You never face a ballot of strangers; you face the shortlist your own base already vetted.
The exact geometry is a dial, not a dogma — how fast the window widens, and how citizens spend their delegation across the tiers: a fixed budget per level, a single pool spent wherever attention goes, or a taper that leans local or national. Those are parameters. The one thing this design proposes to hold fixed is the concept — sight and standing both earned from the base, never handed down from above. And even that is a proposal, not a decree: like every rule here, it is the broadest tier's to keep or rewrite.
The bottom band of that diagram is the load-bearing one. The rules for how the upper tiers are entered and weighted — how delegation is earned, what qualifications matter and how much — belong directly and exclusively to the broadest tier. In the whitepaper's formulation: smaller, specialized cohorts operate within rules they did not write and cannot rewrite, and the broadest cohort's authority over those rules is constitutive, not consultative. Ordinary policy authority may be delegated away; the structure of the delegation itself may not be. That is the answer this design gives to the oldest question in political theory — quis custodiet ipsos custodes?, who guards the guards? Custody of the custodian-rules is held by everyone. Chapter 12 covers how that layer is made slow enough to be safe.