Patients control their FHIR sync window. Maya does the heavy lifting — particle extraction, taxonomy classification, vector embedding, compilation into Compiled Patient Knowledge. This page explains how LeptonX prices that asymmetric work without holding data hostage, and how the Compiled Archive resolves the one-day-extractor scenario at the moment of highest perceived value.
Epic auto-sync allows a patient to authorize a portal connection for as little as a single day. Without a compensating pricing model, a determined user could connect, let Maya ingest and compile a 20-year medical history, then disconnect — absorbing all of our compute cost and walking away with the full compiled output. The pricing architecture below makes that scenario non-extractive by design.
No single lever solves the asymmetric-value problem. The pricing architecture stacks four mechanisms, each defensible in isolation and stronger in combination. The recommended product configuration uses all four simultaneously.
Anyone can connect a portal and Maya will pull the raw FHIR bundle. They see a file tree of what came back. They can download that bundle as a ZIP anytime, for free, forever.
This is the "privacy-first personal data navigator" promise made operational. It is also a legal hedge: HIPAA right-of-access and CCPA portability are satisfied by this single mechanism alone — no further argument required.
Every cancellation flow surfaces the Compiled Archive at $149 one-time ($99 for archives under 5 years; $249 for multi-patient households). Annual subscribers receive it free.
Non-subscribers see it at the moment of highest perceived value — they have just watched Maya compile their record. A meaningful fraction will purchase on the way out. Section 04 details the structure.
Day 1: raw ingest visible, basic structured search. Day 7: Particle Taxonomy®, document rendering, retrieval. Day 30: full NeuroNet personalization, voice query at full fidelity, cross-portal correlation.
This is honest, not artificial. NeuroNet behavior cloning, longitudinal pattern detection, and Portal Intelligence updates genuinely improve with continuous data flow. The soak model aligns user expectation with engineering reality.
Annual subscriptions are aggressively priced relative to monthly ($349/yr vs. $39 × 12 = $468) and include the Compiled Archive bundled quarterly and on cancellation.
A one-day extractor is not paying $349 upfront. The annual tier becomes the structural defense, and the unit economics (Section 05) confirm a single annual subscriber covers the compute cost of 274 worst-case extractors.
These tiers sit atop the eight-tier hardware matrix. They define what the subscription is for — continuous sync, voice query, NeuroNet personalization, and unlimited Compiled Archive exports. Hardware tier determines capability ceiling; subscription tier determines ongoing service.
The Compiled Archive is the universal exit product. Whether a patient subscribes for a decade or a day, they leave with something more organized than they arrived with. Subscribers get it as part of the service. Non-subscribers can purchase it. Every Archive is identical in structure — only the size of the underlying record varies.
A single self-contained directory tree. Folders organized by year, provider, condition, and document type. Rendered PDFs, normalized lab tables, vector-indexed for offline search, with a portable HTML viewer that runs in any browser. LeptonX CheckSum™ integrity manifest included.
No subscription required to read it. No internet required to open it. No proprietary format. The Archive outlives Maya, outlives any subscription, and outlives LeptonX itself.
"We never lock you in. When you leave, you leave with something more organized than you arrived with. That is the privacy-first personal data navigator promise made concrete."— LeptonX Strategic Pricing Reference, May 2026
DGX Spark compute is a sunk capital cost; marginal cost per patient compilation is dominated by electricity and storage. The four scenarios below assume conservative pricing and a fully-loaded patient with a 20-year medical history.
Each guardrail is honest about what it does. None of them hold patient data hostage. They cap compute exposure, gate the most expensive features behind subscription, and prevent rapid re-ingest cycles — while preserving the sovereignty floor at every layer.
Investor-grade transparency. Every element of this pricing architecture carries assumptions that should be tested against beta data, regulatory counsel, and competitive movement. Here is the honest accounting.
The numbers on this page are strategic anchors for the architectural model, not committed pricing. Final pricing will be calibrated against beta cohort feedback, payment processing terms, and the competitive landscape (Apple Health, Validic, Health Gorilla, etc.) at launch.
The economic case assumes 15–30% of departing users will purchase the Compiled Archive. Bracketed by analog products (paid export tools, photo album services, genealogy archives) but unvalidated for medical records. Beta cohort instrumentation will measure this directly.
Marginal cost per patient is dominated by electricity, storage, and bandwidth — not GPU amortization. Per-patient costs in Section 05 assume the Sovereign tier (Tier 1–2) hardware is already owned by the household. Cloud-fallback (Mode 3) economics would change this and remain counsel-gated.
Free raw FHIR export is required by HIPAA, full stop. The Compiled Archive is a value-add product built on top of that legal floor. State privacy law (CCPA portability obligations) is being reviewed by Stacey Gulick at Garfunkel Wild; informal Manatt opinion confirms BAA-exemption.
Every personal-data SaaS faces a version of this. The closest analogues — Apple Health Records, PicnicHealth, Particle Health — have not publicly solved it elegantly. The Compiled Archive model is, to our knowledge, novel in the medical data sovereignty category and may warrant trademark filing.
Thesis 03 (time-locked compilation) and G3 (re-ingest cooldown) require persistent per-patient state across sessions. Straightforward on the patient's own device but requires careful design at account level to avoid violating the zero-cloud-PHI commitment. Local state only.