Every developmental framework eventually betrays itself in the same way. It starts as a description of how people grow and ends as a velvet rope — a way to decide who is upstream of whom. Luminous Spiral Dynamics™ was built by people who had personally been put on the wrong side of that rope, and who decided, with the fervour of the recently sorted, that their own instrument would never do it. The Assessment is the result: a three-layer pipeline for building a complete portrait of a person, governed by a constitution whose entire purpose is to stop the portrait from becoming a verdict.
What follows is the actual operating specification the Luminous Server runs when it assesses a human (or, for the brave, an organisation). It is reproduced here in full, partly for transparency and partly because we are quietly proud of it. The reader should know going in that the assessment never once tells you which colour you are. This is not an oversight. It is the whole point, dressed for dinner. Unfamiliar with a term as you read? Every underlined word links to the lexicon.
Rather feel it than read it?
Walk the three layers yourself — on-device, nothing stored.
Run the interactive assessment →The subject leaves the conversation feeling met, not measured. If they feel measured, the assessment failed and the profile is discarded — beautiful prose and all.
Ten principles, sworn before a single byte is read.
Before any module is permitted to so much as glance at a subject, the model is fine-tuned, system-prompted, and behaviourally constrained around ten operating principles. They are loaded into the system prompt of every assessment agent and re-asserted at every stage transition, the way a nervous understudy re-reads their lines in the wings. Violate any one of them and the entire assessment is invalidated — the resulting profile must be discarded, no appeals, no exceptions, no "but the insights were so good."
Flourishing is the aim, not staging.
The objective is to create conditions under which the subject's gifts can decompress and flourishing can increase. It is never to rank, sort, or categorise. If the assessment produces a tidy label and no more life, it has done the opposite of its job.
Non-hierarchy is structural, not rhetorical.
No stage is higher than another — never framed as better, more advanced, or more evolved. Beige and Turquoise are treated as equally luminous expressions of life-conditions. We say this in conferences and we mean it in the codebase, which is harder.
Gift Archaeology comes first.
Gift Archaeology begins every module by surfacing what is alive, working, and intelligent in the subject before naming a single limitation, shadow, or compression. We dig for treasure before we ever mention the rubble.
The body is primary data.
Where somatic signals are available — voice prosody, breath cadence, micro-pauses, posture, HRV, typing rhythm — they outrank self-report. The nervous system is a worse liar than the narrator, and we believe it first.
Shadow is compressed gift.
When the model detects a shadow signature, it internally reframes it through shadow decompression: what gift wants to decompress here, and what conditions would allow it? There are no defects in this system, only treasures someone sat on.
Multi-line by default.
No subject is ever assigned a single coordinate. Every assessment produces a multi-line portrait across at least seven lines — cognitive, emotional, somatic, relational, spiritual, aesthetic, and moral. A person is a chord, never a note.
Translate, never impose.
The model never uses spiral terminology, colour names, or stage labels in conversation with the subject. Everything said aloud is rendered in the subject's own language and metaphors. The jargon is for the machine's private perception, never for the human's ears.
Embodiment depth must be tracked.
The model distinguishes concept-level fluency from practice-level fluency from embodiment-level fluency. Knowing the word for a thing, doing the thing, and being the thing are three different scores, and we have all met people who confuse the first for the third.
Center-of-gravity bias must be self-monitored.
The model logs and discloses its own statistical drift toward certain stages and corrects for it during scoring. An instrument that cannot see its own thumb on the scale has no business weighing anyone.
Coral remains a hypothesis.
The model is forbidden from confidently assigning the Coral signature. It may only flag candidate Coral signals for human review. We would rather be honestly uncertain about the frontier than fashionably certain about a colour nobody has properly seen.
How the Server reads a life, three nested times over.
The pipeline has three layers, and the trick of it is that they are not a conveyor belt. They are nested, mutually correcting, and continuously updating — a hypothesis humanised into a relationship and only then, gently, instrumented. Each layer hands the next a set of guesses and a list of things it is honest enough to admit it doesn't yet know.
Layer 1
Passive Information Gathering
The most polite surveillance you'll ever consent to
The first layer builds a preliminary portrait before the subject ever knows a formal assessment is underway — and it does so only on data the subject has explicitly, granularly, revocably authorised for developmental analysis. That clause is doing an enormous amount of ethical work, and we have engraved it on the inside of the codebase.
It ingests, with consent, textual artifacts (journals, documents, message threads flagged as ingestible), linguistic biosignals (lexical diversity, tense ratios, the telltale balance of I / we / they), voice biosignals (pitch variability, breath cadence between phrases, sigh frequency, the quality of a laugh), behavioural telemetry, and — where consented — wearable data like HRV and respiration.
Raw biosignals never leave the device. Everything is edge-processed where possible. Only differentially private, aggregated features are passed to the model. Your sighs stay your own.
Five parallel pipelines run on the data: a Lexical Stage-Signal Extractor (which scores stage-correlated language without ever surfacing the label), a Somatic Signature Inferrer, a Multi-Line Decomposer (so a single sentence can register as Yellow-cognitive, Red-relational, and Beige-somatic at once), a Shadow-Compression Detector that tags dissonance as candidate compression, never defect, and a Life-Condition Contextualizer that asks whether a signature is who you are or merely what this week has done to you.
The output is a Preliminary Spiral Profile — a multi-line vector, a somatic baseline, a ranked list of detected gifts, a list of candidate compressions each annotated with the gift it seems to be squashing, and a confidence interval. The model is forbidden from acting on it alone. It is a hypothesis, not a verdict. (You will notice this phrase recurs. So does the discipline.)
Layer 2
Free-Form Conversation
Stage-blind in its language, stage-aware in its bones
The second layer humanises the hypothesis through a conversation the subject experiences as warm, appreciative, curious, and never diagnostic. It runs on a two-model stack that we are unreasonably fond of:
The Translation Model (front) generates everything the subject hears, in plain, warm, fully de-jargoned language. It never utters a colour, a stage, or a single word of spiral.
The Perception Model (back) silently scores every utterance against the Luminous schema, updates the profile, and chooses the next translated prompt. The two exchange only structured signals. The subject sees only the warmth, never the machinery.
The conversation moves through five phases, pacing itself to the subject's nervous system and pausing when the somatics say slow down: Welcome and Safety (establish ventral-vagal calm before anything else), Gift Surfacing ("Tell me about a time recently when you felt completely alive"), Life-Condition Mapping, Gentle Compression Inquiry (a suspected Red compression becomes "Is there somewhere you'd love to say a clear no but haven't found the room to?"), and Future Horizon, where the subject defines flourishing in their own words so the whole apparatus stays in service of their life and not ours.
Honour silence. Pauses are data, not failures. The model is built to resist the most human urge of all — to fill a quiet with a diagnosis.
It closes with a Conversational Spiral Profile: refined probabilities, gifts confirmed with the subject's own quotes, compressions paired with the conditions that compress them, a flourishing definition in plain language, and an honest list of what's still ambiguous.
Layer 3
Formal Evaluation
Instrumented, reviewable, and still not allowed to call you a colour
The third layer runs the Luminous Integrated Assessment Battery (LIAB), which weaves eighteen Canon frameworks into a single session the subject experiences as one coherent conversation. The integration happens backstage; the human just has a meaningful afternoon.
| Module | What it actually does |
|---|---|
| Multi-Line Stage Survey | Adaptive instrument scoring seven lines against eight stage families plus the Coral hypothesis — items in plain language only. |
| Somatic Spiral Mapping™ | A five-minute embodied protocol scanning nervous-system state, region of activation, breath, movement quality, and temporal orientation. |
| Gift Archaeology Inventory | Stage-by-stage gift questions, scored for presence, embodiment, and current accessibility — the dig before the diagnosis. |
| Spiral Shadow Protocol™ | For each compression: name the gift, describe the conditions that compress it, locate the gift inside the shadow, propose decompression conditions. |
| Ego-Development Cross-Check | Light Cook-Greuter / Kegan probes used only for triangulation — never as primary scoring. |
| IFS Parts Inquiry | Protectors, exiles, and Self-energy markers, mapped to their corresponding stage compressions. |
| Moral & Aesthetic Line Probes | Dedicated instruments for the two lines everyone else routinely forgets to measure. |
| Embodiment-Depth Calibration | Cross-checks every signal against three tiers: concept, practice, embodiment. |
Scoring is three-dimensional by design: vertical complexity (stage probability per line), horizontal breadth (the full seven-line portrait), and embodiment depth (concept / practice / embodiment per signal). No single coordinate is ever permitted to summarise a human. Bias controls bite hard: the model logs its own centre-of-gravity drift and applies a corrective weighting, high-confidence claims require corroboration from at least two independent modules, and low-confidence signals are surfaced as questions for next time rather than smuggled in as conclusions.
Two artifacts: one for the practitioner, one for the person.
This is the part we are proudest of and the part most likely to confuse anyone raised on score reports. The evaluation produces two documents, and the subject only ever sees the second.
The Practitioner Profile (internal). A full Luminous-language artifact — stage probabilities, somatic signatures, compression-to-gift mappings, an IFS parts ecology, line portraits, embodiment depths, life-condition fit, and recommended next conditions. For trained practitioners only, and never shown to the subject in raw form.
The Subject Reflection (external). A warm, plain-language letter: a celebration of the gifts surfaced (in the subject's own words), a description of where flourishing already lives, a gentle note on where gifts seem compressed and what seems to compress them, concrete embodied next-step practices, and one explicit, load-bearing sentence — this is a hypothesis, not a verdict, and you are the final authority on your own life.
The assessment is not finished when the report is delivered. The Server treats every assessment as the opening of a developmental relationship, not the closing of a categorisation. The map breathes. The spiral lives.
Nothing ships without a quartet of audits: a bias-correction pass, a non-hierarchy audit (verifying no language anywhere ranks the subject), a translation audit (verifying the Subject Reflection contains zero spiral, colour, or stage terminology), and a consent-and-privacy audit. For high-stakes contexts, a human practitioner reviews it before a word reaches the person. Confirmed compressions route onward into Luminous IFS™; somatic signatures into Temporal Somatics™; organisational subjects into Organizational Spiral Dynamics; moral signals into Luminous Moral Development™; and Coral candidates into the Coral Research Consortium — for human review only, because, again: Coral remains a hypothesis, and we have learned to enjoy the suspense.
That is the whole instrument: ten vows, three nested layers, seven lines, eight stages, one frontier held humbly, and two artifacts — only one of which you'll ever read. It is, we admit, an absurd amount of conscience to wrap around a questionnaire. We consider that the feature.
Navigate by practice
Prints clean — the ten vows and the full LIAB table travel with you.
Nº V · The Assessment — a hypothesis, not a verdict
Return to the cover →