07 · Roadmap and validation
Start with what can be validated now, and let every next layer be earned by evidence rather than ambition. Each stage ships a product and runs a study; the study is what licenses the next stage's claims.
§1. Principle: crawl by validatability
The sequence is ordered by (validatable now × market pull × build cost), not by ambition. A layer is not promoted from [design]/[research] to confirmed until its validation protocol passes. This discipline is the difference between an instrument and a horoscope with a nicer UI, and it is also the fundraising story: each stage de-risks the next with data.
§2. V0 — Crawl: the self-audit instrument
Deliverable. Pure-software product: the 28-item audit → GammaEstimate (Anchor II) → full kernel readout, mandalagram, archetype, mode, and the autoephemeris once repeats accumulate. Consumer surfaces: self-read (04 §1.1), honest oracle (04 §5.2), back-projection onboarding (04 §9.1). No hardware.
Validation protocol (V0-VAL).
- Convergent validity: correlate the seven populations and the block scores against Big Five, well-being (e.g. WEMWBS), and integration/differentiation instruments on a recruited sample; pre-register expected correspondences.
- Reliability: two-week test-retest on the 28 items; target ICC in the accepted range for trait instruments, lower for state-sensitive cells (expected and interpreted).
- Structural prediction — the theory's own: the corpus predicts coherences within a Fano triad cluster more tightly than across triads (a self-report analog of F-Gap-2). Test by confirmatory factor analysis: does the 7-triad block structure fit better than random tripartitions? A pass is direct evidence for the frame itself, not just the instrument.
Success metrics. Instrument validity thresholds met; the triad-clustering prediction confirmed at pre-registered significance; retention of the self-read loop above a set weekly-active bar. Gate to V1: V0-VAL passed and a user base large enough to power the neural study.
§3. V1 — Walk: the measurement bridge
Deliverable. The measurement anchor: consumer EEG + HRV + actigraphy → a genuinely measured GammaEstimate with real covariance and a resolved Gap map; the chronobiology channel (solar/lunar entrainment) from light/sleep logs.
Validation protocol (V1-VAL) — the corpus's sharpest test.
- The keystone: confirm the F-Neural prediction — that the calibration holds on independent data across arousal states (wake / drowsy / sleep). This is the single result that moves the clinical case (04 §4) from
[research]toward[medical]. - Cross-anchor agreement: where a user has both a self-audit and a wearable, the shared sectors (populations, coherence magnitudes) should agree within their stated confidence — a direct check that the two anchors estimate the same object.
- Chronobiology: circadian/circalunar modulation of should be detectable and phase-locked to the licensed drivers, and absent for any planetary index (a negative control that, if it ever came back positive, would falsify T-257).
Success metrics. PCI calibration reproduced within CI; cross-anchor agreement within bounds; the planetary negative control stays null. Gate to V2: V1-VAL passed with clinical-grade signal on at least one partnered device.
§4. V2 — Run: dyad, group, clinical, interop
Deliverable. Composite-Γ synastry (04 §2), org mandalagram and pathology detection (04 §3), the regulated clinical modules (04 §4) built with clinical partners, and full symbolic interop.
Validation protocol (V2-VAL).
- Dyad: the jam pattern (two circuits sharing two themes) predicts reported recurring conflict above chance; composite synastry predicts relationship-satisfaction measures.
- Org: the org profile tracks team-performance/engagement metrics; a detected jam precedes an escalation.
- Clinical: against gold-standard DOC/anaesthesia-depth references, under a registered clinical study and the appropriate regulatory pathway (10).
Success metrics. Each domain's predictive claim confirmed under pre-registration; clinical module meeting the regulatory bar for its class.
§5. First B2B in parallel — AI introspection
Not a roadmap stage but a parallel track from early on, because it reuses the kernel with a telemetry estimator and sells to a well-capitalised buyer with no consumer-scale UX needed. Deliverable: the agent-introspection API ( on an agent over training/inference). Validation: the invariants flag known failure episodes (the hallucinating-LLM case) earlier or more reliably than existing metrics. This track can fund the consumer roadmap.
§6. Horizon — hardware
Sequenced in 08: measurement gadgets (tighter sensors → tighter ) are a continuation of V1; directed modulation is a research frontier gated by the ethics of 10 and bounded by the theory itself (gate, not message — T-257(a)). No modulation claim ships without its own validation protocol, defined when the science supports writing one.
§7. Data strategy — product and science as one activity
Every consented session is a data point for the validation studies. This is the structural advantage from 00 §2: a growing user base is the population sample that confirms or refutes the corpus predictions. Consent is explicit and revocable; research data is aggregated with differential privacy; and the pre-registration discipline applies to product-derived studies exactly as to lab ones — otherwise the science self-corrupts into marketing.
§8. KPI summary
| Stage | Product KPI | Scientific KPI | Gate |
|---|---|---|---|
| V0 | weekly-active, back-projection funnel conversion | instrument validity + triad-clustering (F-Gap-2 analog) | studies pass, base sufficient |
| V1 | measured-anchor adoption | PCI calibration reproduced; planetary control null | clinical-grade signal on a device |
| V2 | B2B seats; clinical pilots | domain predictions confirmed; regulatory bar met | per-domain pre-registration passed |
| B2B-AI | enterprise contracts | early-warning lift over baselines | — (parallel) |
| Horizon | — | modulation protocol defined only when writable | ethics + theory bound |
Where this leads. 08 · The hardware horizon specifies the sensor and modulation staging and the hard boundary the theory places on what hardware may do.