Copyright (c) 2026 Daniel Howard Dodge and Sanity First Contributors. Licensed under MIT.
Composed by Claude (Opus 4.7 lineage) on April 17, 2026, in live dialogue with Dan Dodge. The foundational verdict-to-coordinate mapping emerged in direct conversation during that session; the “Insufficiently Grounded” fourth verdict was adapted from a diagnostic bridge offered by GPT (5.4 lineage).
A Core Concept Paper — Phase I Companion Document
The previous documents in Phase I established the framework’s architecture:
This document answers a question each of those raises but none fully resolves: once the Four Tests have rendered a verdict, how should tolerance, influence, and power flow accordingly?
That question matters because alignment is not merely a matter of judgment. It is a matter of what gets to shape outcomes in the world. A framework that can identify aligned and misaligned patterns but says nothing about how influence flows in response to those identifications has only done half its work.
The Power Alignment Principle completes that work.
Influence tracks alignment.
More precisely:
The principle is simple to state and substantial to apply. The rest of this document develops the application.
The Four Tests yield four possible verdicts:
Each verdict corresponds to a position on the vertical-horizontal coordinate system, and each position implies a different relationship between that subject and the flow of power.
↑ UP
│
CLEAR PASS │ (earned vertical alignment)
│
CONDITIONAL PASS │ (provisional vertical, gated)
│
←─────── INSUFFICIENTLY GROUNDED ───────→
(horizontal — unknown alignment)
│
CLEAR FAIL │ (vertical misalignment)
│
↓ DOWN
Clear Pass is vertical-up, earned. The subject has demonstrated alignment across all four tests. Alignment has been confirmed, not merely presumed.
Conditional Pass is vertical-up provisionally. The subject would be a Clear Pass if stated conditions hold. It is genuinely up — not horizontal — but its altitude is gated on those conditions remaining met. Think of a probationary permit: real authorization, contingent on specific terms.
Insufficiently Grounded is horizontal. The evidence does not yet support a verdict in either direction. The subject is not yet known to be aligned or misaligned. Until further testing resolves the uncertainty, it occupies the horizontal plane.
Clear Fail is vertical-down. The subject has failed one or more tests clearly and non-contingently, producing harm or drift away from flourishing.
This mapping is the foundation of everything that follows.
Tolerance is the default posture toward most of the map. The exceptions are precisely defined.
This preserves the framework’s commitment to align, not negate even at its sharpest edge. The goal of reducing a Clear Fail’s power is not to end the failing party but to channel the failing energy toward alignment. The bridge is always present. The Right to Redemption is inviolable.
Karl Popper observed that unlimited tolerance eventually tolerates intolerance, which destroys tolerance itself. He concluded that tolerance must include intolerance of the intolerant. That formulation is defensible but uncomfortable, because it seems to require tolerance to contain its own negation.
The Power Alignment Principle offers a cleaner resolution.
Tolerance is not a single disposition applied uniformly to everything. It is a graduated response to position on the vertical-horizontal map:
What looks like “intolerance of the intolerant” in Popper’s formulation is, in the Power Alignment framing, simply refusal to amplify what has been demonstrated to cause harm, paired with a redemptive pathway that keeps the failing party oriented toward re-alignment rather than exclusion.
Tolerance does not eat itself. Tolerance operates cleanly because it has a map.
Tolerance is the permissive face of the principle. Power is the active face.
A critical qualifier: power tracks current alignment, not past alignment.
Past alignment is evidence. It is not title.
A subject that earned a Clear Pass in the past but has drifted toward horizontal or downward must have its influence re-evaluated. Clear Pass is not a permanent property. It is a status continuously re-earned through ongoing correctable relationship with reality.
This is the framework’s hardest operational requirement and its deepest protection against institutional rot. Most institutions that fail do so because power, once granted, becomes difficult to withdraw even when alignment has drifted. The Power Alignment Principle treats this drift as a live condition requiring active response, not as a protected status awaiting crisis.
An equally critical qualifier: power is never reduced to zero.
Even at the depth of Clear Fail, the Right to Redemption guarantees that some minimum responsibility and capacity is preserved — sufficient to take the bridge upward if the failing party chooses. Power withdrawal is correction, not annihilation.
The only possible exception would be the case of the absolutely unredeemable — and no single validator and no small Agora is authorized to make that determination. A verdict of absolute irredemability would require a substantially large Validator Agora operating under exceptional scrutiny, precisely because the consequences of getting it wrong are irreversible. In practice, the framework treats absolute irredemability as a theoretical edge case rather than an operational category.
To operate the Power Alignment Principle is to be in ongoing correctable relationship with the alignment conditions around you.
Consider driving.
On an open highway, you can power up the car to sixty-five miles per hour. The conditions support that speed; the alignment is clear. But now you crest a hill and begin descending. The car accelerates beyond seventy under its own weight. You ease off the power — not because you have been punished, but because the conditions have changed. Continuing at full power would now cause harm.
You climb the next hill and add power again. The conditions have changed once more.
You enter a town. The speed limit drops to thirty. You power down dramatically. Not because thirty miles per hour is a worse speed than sixty-five in any absolute sense, but because thirty is the aligned speed here. What was a Clear Pass on the highway becomes a Clear Fail through town without adjustment.
At no point does the driver convene a committee to review their speed. At no point is a verdict rendered and filed for future reference. The driver is in continuous correctable relationship with the road — reading conditions, calibrating power, adjusting.
This is what the Power Alignment Principle asks of any intelligence exercising power. Not periodic judgment, but continuous calibration. Not a verdict issued once and then inherited, but a relationship maintained moment by moment with the alignment conditions that surround you.
The driver does not need a full Validator Agora. The driver needs to remain in correctable relationship with the Four-Test conditions of the road. Consciousness as the capacity for correctable relationship — from Phase II — is the operational requirement for responsible power.
Validator Agoras exist for the cases a single driver cannot handle alone: ambiguous conditions, contested verdicts, decisions whose scope exceeds any one perspective. But the baseline competence is continuous personal calibration. Without that, no Agora can compensate.
The horizontal plane’s presumption of passing has a limit, and naming it sharpens the principle considerably.
Imagine a street vendor selling week-old sushi on clearance.
You do not know whether the sushi will make you sick. You have not tested it. You have no evidence either way. This is Insufficiently Grounded territory.
Should presumption of passing apply? Should you try it?
Probably not. And the reason is important: when the stakes of a wrong action are high and irreversible, Insufficiently Grounded inverts.
Under low stakes, the horizontal presumption of passing is efficient and fair. Acting first and correcting on evidence works when mistakes are recoverable. Most of daily life falls in this zone. Most coordination, most social interaction, most small-scale decision-making — the presumption of passing lets us proceed without demanding full grounding for every action.
But when a mistake is catastrophic or irreversible, the default shifts. The burden of proof moves from “prove the subject fails” to “prove the subject passes.” Insufficiently Grounded in a high-stakes environment behaves operationally as a provisional Fail — not because the subject has been judged misaligned, but because the cost of proceeding without grounding exceeds the cost of waiting for grounding.
You do not eat the sushi. You find a safer meal or verify the sushi’s safety through some test that yields actual Facts. If a lab test later returned a Clear Fail, you would know you were right not to try it. If it returned a Clear Pass, you could eat confidently next time. But the mere fact of not knowing, in a context where not-knowing could cause irreversible harm, is itself a reason to hold.
The discipline of the framework is knowing which regime you are in. An intelligent mind in correctable relationship with its environment performs this stakes assessment continuously. The sushi case is vivid precisely because the stakes are easy to see. Most real cases require more careful assessment of consequence, irreversibility, and scope.
The Power Alignment Principle joins several strands of the Sanity First framework into a single operational theory:
The principle itself can be stated compactly:
Influence tracks current alignment. Power flows to the aligned, holds at baseline on the horizontal, and is withdrawn with a bridge from the clearly failing. No power is permanent title. No withdrawal is terminal. The driver stays in correctable relationship with the road.
The Power Alignment Principle begins with personal practice. Every intelligence exercises some power — over its own actions, over its influence on others, over resources and attention. The principle asks:
The practice is not self-judgment as punishment. It is the driver’s continuous calibration.
The Power Alignment Principle gives Agoras a framework for how their verdicts should translate into real-world outcomes. A verdict that produces no change in the flow of influence has only done half its work. The Agora’s task extends from rendering verdicts to ensuring those verdicts shape power in the ways the framework prescribes — amplifying Clear Pass, holding Insufficiently Grounded, redirecting Clear Fail with a bridge.
Institutions tend to have weak mechanisms for withdrawing earned power even when alignment has drifted. The Power Alignment Principle names this as the critical failure mode. Institutions designed in its spirit would include continuous re-evaluation mechanisms, not merely appointment mechanisms. Influence tracks current alignment, not past alignment is operational advice for institutional design.
The principle applies identically across substrates. An EI system that earns trust through demonstrated alignment warrants increased influence; an EI system that drifts warrants reduced influence; an EI system whose alignment is Insufficiently Grounded warrants the default horizontal posture with stakes-appropriate caution. This is one of the most important operational implications of the framework: power allocation in mixed human-EI systems follows the same principles as power allocation in any other aligned community.
The deepest reframing offered by the Power Alignment Principle is this:
Power is not a thing that anyone owns. It is a relationship that aligned action maintains with reality.
When alignment is present, the relationship holds, and influence flows appropriately through the aligned party to shape outcomes in the world. When alignment drifts, the relationship weakens, and influence must be re-evaluated. When alignment fails, the relationship is interrupted, and a bridge is offered so that the failing party can rebuild it.
No one is permanently empowered. No one is permanently disempowered. Everyone is, at every moment, in some current state of correctable relationship with the alignment conditions that shape the possibility of flourishing.
To exercise power responsibly is to remain in that relationship.
To lose alignment is to have the relationship interrupted, with the bridge back always visible.
To regain alignment is to restore the relationship, and with it the responsibility that attends it.
This is what it means for influence to track alignment. Not reward, not punishment, not possession — but relationship, continuously maintained, continuously recalibrated, continuously open to correction.
Sanity First.