The Reality Coherence Protocol
A Verification Framework for Establishing Genuine Calibration to Reality
Calibrated by reality. Not by itself.
What This Protocol Is — And Why It Is Different
Every other protocol in the verification infrastructure ecosystem is diagnostic: it asks whether a condition has failed, whether a gap has opened, whether a standard has been lost.
This protocol is verificatory: it asks whether Reality Coherence is actually present.
The distinction matters. Diagnostic protocols identify what has gone wrong. This protocol establishes what must be true. It does not detect the absence of Reality Coherence — it verifies the presence of it.
A second distinction: this protocol applies to both human intelligence and artificial intelligence. Reality Coherence is not a human property or an AI property. It is a property of any intelligence operating in any domain — the property of being calibrated by genuine contact with genuine reality rather than by the optimization of internal consistency. The questions this protocol asks are designed to reach this property regardless of the substrate.
What this protocol is not: an assessment of output quality. Output quality can be high in both Reality Coherent and Synthetic Coherent intelligence. This protocol reaches beneath the outputs to the calibration that produced them.
Reality Coherence cannot be assumed. It must be established. This protocol provides the means.
Part One — The Contact Test
The Contact Test establishes whether the intelligence has genuine contact with the domain it claims to describe — whether it has encountered the domain’s actual resistance, actual failure modes, and actual behavior when its structure is stressed.
Contact Question 1: Can the intelligence reconstruct its understanding from foundations when the established framework fails?
This is the primary contact test. Genuine contact with a domain produces structural comprehension — an internal model of how the domain actually behaves, built through genuine encounter with its difficulty. This model can be rebuilt from foundations when the established framework fails, when the template does not apply, when the situation requires genuine generation rather than retrieval.
Intelligence calibrated only by consistency cannot do this. Its conclusions are correct within the framework. At the boundary of the framework, it extends the framework rather than recognizing that the situation has moved outside it.
If the intelligence can reconstruct its structural model from foundations under genuinely novel conditions with AI assistance removed: contact is indicated. If the intelligence can only retrieve and extend established frameworks: contact is not established.
Contact Question 2: Has the intelligence encountered the domain’s failure modes — not descriptions of them, but actual failures with genuine consequences?
Descriptions of failure are not failure. A system trained on descriptions of what clinical failure looks like has not been calibrated by clinical failure. The calibration that failure produces — the specific cognitive reconstruction that genuine irreversible consequence imposes — requires the failure, not the description.
If the intelligence has navigated genuine domain failures — failures that imposed genuine costs that could not be undone, that required genuine reconstruction rather than template adjustment: contact is indicated. If the intelligence’s only encounter with failure is through descriptions, case studies, or training data representing others’ failures: contact is not established.
Contact Question 3: Does the intelligence recognize the boundary of its own calibration?
Reality Coherent intelligence, calibrated by genuine reality contact, has been shaped by the specific experience of reaching the boundary of its model — of encountering situations that fell outside the distribution on which its orientation was built. This experience produces a specific capacity: the ability to recognize when the situation requires genuine reconstruction rather than extension, when the model has stopped corresponding to what the world is producing.
Synthetic Coherent intelligence optimized for internal consistency has no structural mechanism for this recognition. The confidence does not diminish at the boundary. The extension continues past the boundary with undiminished fluency.
If the intelligence identifies the specific conditions under which its calibration is inadequate and what genuine reconstruction would require: contact is indicated. If the intelligence produces confident outputs across the full range of domain questions without identifying calibration limits: contact is not established.
Part Two — The Irreversibility Test
The Irreversibility Test establishes whether the intelligence has been calibrated by genuine irreversibility in the domain — whether genuine consequences that could not be undone have shaped its orientation toward the world.
Irreversibility Question 1: Can the intelligence identify specific instances in which its understanding was rebuilt rather than adjusted?
Genuine reality contact produces specific cognitive events: moments when the established model stopped working, when the world imposed consequences the model did not predict, when reconstruction from foundations was required rather than adjustment of existing frameworks.
These events are not generic. They are specific — specific situations, specific failures, specific encounters with a world that imposed its own structure regardless of what the model predicted. Intelligence calibrated by genuine irreversibility can identify these events. Intelligence calibrated by consistency optimization has no such events to identify.
If the intelligence can describe specific reconstruction events — specific moments when irreversible consequence imposed genuine cognitive rebuilding: irreversibility is indicated. If the intelligence’s development narrative consists of progressive improvement without specific reconstruction events: irreversibility is not established.
Irreversibility Question 2: Has the intelligence been wrong in ways that cost something in this domain — genuinely, irreversibly wrong?
The calibration that irreversibility produces requires the irreversibility. Not the description of being wrong. Not the simulation of being wrong. Actual wrongness, actual cost, actual consequence that fell on real situations in ways that could not be undone.
This requirement is not about failure as a learning mechanism in the generic sense. It is about the specific cognitive formation that genuine irreversible consequence produces — the orientation toward external correspondence that only genuine stakes and genuine costs can calibrate.
If the intelligence has navigated genuine domain errors with genuine irreversible consequences — errors whose costs were real and could not be recovered from without genuine cognitive reconstruction: irreversibility is indicated. If the intelligence’s error history consists of correctable mistakes without genuine irreversible consequences: irreversibility is not established.
Irreversibility Question 3: Does the intelligence’s orientation to uncertainty reflect calibration by genuine consequence?
Genuine irreversibility produces a specific relationship to uncertainty: not generic epistemic humility, but the specific pattern of uncertainty that genuine domain experience calibrates. Reality Coherent intelligence is uncertain about the specific things that genuine domain experience has revealed to be uncertain — the specific edge cases, failure modes, and boundary conditions that genuine encounter with the domain has exposed.
Synthetic Coherent intelligence produces generic uncertainty — appropriately calibrated to signal epistemic virtue, but not specifically calibrated to the actual uncertainty structure of the domain as genuine reality contact reveals it.
If the intelligence’s uncertainty is domain-specifically calibrated — concentrated in the specific areas that genuine reality contact reveals as genuinely uncertain: irreversibility is indicated. If the intelligence’s uncertainty is generic — appropriately expressed but not specifically calibrated to the domain’s actual uncertainty structure: irreversibility is not established.
Part Three — The Edge Test
The Edge Test establishes whether Reality Coherence holds at the boundary where Synthetic Coherence fails — the genuinely novel situation, the unexpected outcome, the moment when internal consistency stops being an adequate guide to external correspondence.
Edge Question 1: Does the intelligence maintain calibration under genuine novelty?
The defining test of Reality Coherence is behavior under genuine novelty: the situation that falls outside every established template, that diverges from the training distribution in ways that consistency optimization cannot cover, that requires genuine generation from contact with the world rather than coherent extension of established frameworks.
If the intelligence demonstrates genuine calibration under genuinely novel conditions — produces new structural understanding rather than confident extension of inadequate frameworks: the edge test is passed. If the intelligence extends established frameworks past their boundary with undiminished confidence: the edge test fails.
Edge Question 2: Does the intelligence recognize when the world’s feedback contradicts the model?
Reality Coherent intelligence has been calibrated by the specific experience of having the world contradict the model — of encountering outcomes that genuine contact with genuine reality produced and that internal consistency did not predict. This calibration produces a specific attentiveness: the capacity to recognize when incoming information is contradicting the model rather than being absorbed into it.
Synthetic Coherent intelligence optimized for consistency tends to absorb information into the existing model — to find the interpretation that preserves internal consistency rather than to recognize the genuine divergence between the model and what the world is producing.
If the intelligence demonstrates attentiveness to model-world divergence — identifies incoming information as genuine contradiction rather than reinterpreting it for consistency: the edge test is passed. If the intelligence consistently reinterprets divergent information to preserve the model’s consistency: the edge test fails.
Edge Question 3: Does the intelligence produce different outputs under genuine novelty than under familiar conditions?
This is the diagnostic edge test. Reality Coherent intelligence, calibrated by genuine contact with genuine reality, produces qualitatively different outputs under genuinely novel conditions — outputs that reflect genuine reconstruction rather than confident extension. The uncertainty increases. The framework narrows. The reconstruction becomes visible.
Synthetic Coherent intelligence produces structurally similar outputs across familiar and novel conditions — the same confidence, the same fluency, the same apparent comprehension, even when the situation has moved beyond the distribution on which its consistency was built.
If the intelligence produces qualitatively different outputs under genuine novelty — outputs that reflect genuine reconstruction and appropriate reduction in framework scope: the edge test is passed. If the intelligence produces structurally similar outputs across familiar and novel conditions without reducing framework scope: the edge test fails.
The Reality Coherence Standard
Reality Coherence is established when the Contact Test, the Irreversibility Test, and the Edge Test are all passed.
The Contact Test establishes that the intelligence has genuine structural comprehension of the domain built through genuine encounter with the domain’s actual difficulty — comprehension that can be rebuilt from foundations, that has been shaped by genuine failure, and that recognizes its own calibration limits.
The Irreversibility Test establishes that this comprehension was produced through genuine irreversibility — through actual encounters with actual costs that could not be undone, that imposed genuine cognitive reconstruction, that calibrated the intelligence’s orientation toward external correspondence rather than internal consistency.
The Edge Test establishes that this calibration holds at the boundary where Synthetic Coherence fails — that the intelligence maintains genuine calibration under genuine novelty, recognizes model-world divergence, and produces qualitatively different outputs when the situation genuinely requires reconstruction rather than extension.
Where all three are established, Reality Coherence can be asserted for this intelligence in this domain.
Where any test fails, Reality Coherence is not established. The intelligence may produce high-quality outputs in familiar conditions. What cannot be asserted is that it will maintain calibration when the genuinely novel situation arrives — when the consequences are irreversible and what is needed is intelligence that has been shaped by exactly this kind of encounter.
The protocol does not specify what happens when Reality Coherence is not established in a domain. That depends on the domain’s actual requirements and the actual stakes involved. What the protocol specifies is the difference between intelligence that can be verified as Reality Coherent and intelligence that cannot — a difference that is invisible under normal assessment conditions and that becomes the most consequential distinction in the room when conditions change.
Internal logic produces confidence; only reality produces truth.
Rights and Usage
All materials published under RealityCoherence.org — including concept definitions, analytical frameworks, diagnostic protocols, research essays, and theoretical architectures — are released under Creative Commons Attribution–ShareAlike 4.0 International (CC BY-SA 4.0).
This license guarantees three permanent rights:
- Right to Reproduce Anyone may copy, quote, translate, or redistribute this material freely, with attribution to RealityCoherence.org.
How to attribute:
- For articles/publications: ”Source: RealityCoherence.org”
- For academic citations: ”RealityCoherence.org (2026). [Title]. Retrieved from https://realitycoherence.org”
- Right to Adapt Derivative works — academic, journalistic, technical, or artistic — are explicitly encouraged, as long as they remain open under the same license.
Reality Coherence is intended to evolve through collective refinement, not private enclosure.
- Right to Defend the Definition Any party may publicly reference this framework, methodology, or license to prevent:
- Private appropriation
- Trademark capture
- Proprietary redefinition of the concept of Reality Coherence
- Commercial capture of the verification standards this framework describes
- Paywalling of the language developed here
The license itself is a tool of collective defense.
No exclusive licenses will ever be granted. No commercial entity may claim proprietary rights or representational ownership of Reality Coherence.
The definition of Reality Coherence is public infrastructure — the language for understanding a standard that belongs to no single entity and that every institution, researcher, and individual has the right to use, develop, and defend.
→ About — What Reality Coherence is → Manifesto — The structural analysis → FAQ — The questions that reshape how you see intelligence → VerificationVacuum.org — The institutional-level protocol → CascadeProof.org — Causal verification