LymeHQ · How it works

One loop, and you hold every gate on it

What you put in, who can see it, what comes back, and — said as plainly — which parts of that are built today and which are not.

Read this first

LymeHQ is in demo. Everything described below runs, and it runs on invented data — every account is synthetic and no real health information is held. The point of building it this way is that nothing about the design changes when that stops being true.

The loop closes

Three stations, one direction — and what the commons learns comes back.

How data moves from patient to provider to researcher, and back to the patientData moves in one direction through three stations. A patient contributes symptoms and labs. A provider sees data from a patient only where that patient has both added them and turned provider sharing on. Researchers reach the data only as aggregates covering five or more people. The loop then closes: group figures from the commons come back to the patients whose own records match the group, as figures only. Every step requires consent, and every step is written to an audit log.PatientSymptomsand labsProviderWhat youshareResearcherFive ormoreFindings return asfigures — group figures only
  • This path works today

Every step needs consent. Every step is written to an audit log.

The return leg carries group figures only — never a comparison between treatments, and never a number about you.

Findings return as group figures. Your own numbers are never shown beside them, and nothing is sent to you.

Orientation only — nothing on this drawing is a measurement. Every claim it makes is one we test on every build, not one we are asking you to take on trust: consent changes are recorded by the database itself, not by the app; a clinician reaches your record only when you have both added them and turned sharing on, and both must be true; every look at health information is written to a log that cannot be edited or erased afterwards; and research results are released only for groups of five people or more. The return leg is drawn solid because it works today. It carries group figures only — a finding may report what a group looked like, never what a treatment did.

01 · The loop

Six stations, in order

Each one says whether it exists today.

  1. 01 · YouBuilt

    A record you keep, in a form a clinician can read

    A daily check-in that takes under a minute, the medications you are taking and when you started them, lab results you upload, and the people and animals in your household. Nothing is required and nothing is scored — a gap in the record is a day you were unwell, not a failure, so nothing here counts consecutive days at you.

    The dates come from the server rather than from a form, so a timeline cannot be quietly rewritten after the fact. That matters more than it sounds: a symptom history someone could edit is not usable in front of a clinician, and it is not usable as research either.

  2. 02 · Making sense of itBuilt

    Arithmetic first, language second

    Whether your symptoms moved after a treatment started is decided by counting, not by a model. One piece of code computes the direction, the size and whether there is enough data to say anything at all — and when there is not, it stops before anything is generated.

    Beacon can then phrase the result, but it cannot invent one. Telling a chronically ill person they are improving when the data cannot say is the failure this design exists to make impossible, so "not enough data yet" is a row count rather than a model’s confidence.

  3. 03 · Your clinicianBuilt

    They see what you share, from the moment you share it

    You add a clinician; you switch sharing on. Both are yours to do and yours to undo, and the second you undo either one their view is gone on their next page load. What they see is your timeline and the figures around each treatment you have recorded — never your private journal.

    There is nothing on LymeHQ that lets a clinician request access, claim a patient, or be assigned one. The only way a record is shared is that you shared it.

  4. 04 · ResearchBuilt

    Group counts, never records

    If you opt in, your data joins a pre-computed table of group counts — how many people match a combination of characteristics. Researchers query that table and nothing else. No combination describing fewer than five people is ever written into it.

    They can find out whether a cohort exists before applying for anything, because forcing an application first would produce applications written blind. Getting the actual figures needs an approved research request for a specific dataset, and every query is recorded.

  5. 05 · Your householdBuilt

    A dog’s routine test is an early warning for the people around it

    Dogs are tested for tick-borne exposure yearly, stay close to home and cannot hide symptoms — which makes a positive result a useful statement about where ticks are. Record one and everyone in the household is told, once.

    The alert says where ticks are, never what anyone has, and no sharing switch gates it. It is your household’s own safety signal, and making it conditional on contributing pet data to the commons would be charging for it.

  6. 06 · Back to youBuilt

    What research learns, returned as figures

    Findings from the commons come back to the people who made them possible, as figures beside the same guarded language everything else uses. Nothing is sent to you: the page matches your own record against each group when you open it, and stores nothing about which ones matched.

    One decision about it is already fixed and will not change: the return path is figures only. A causal sentence — "your fatigue eased after you started X" — is never generated for a patient, at any milestone.

02 · Beacon

A companion that cannot diagnose you

Not because it is instructed not to — because of how it is wired.

Beacon answers questions about Lyme and tick-borne illness. Whether it can see your record is a switch you hold, and it is off until you turn it on; with it off Beacon still works and simply knows nothing about you. Turning it off takes effect on your next message, with no cached context carried over.

Two things are structural rather than instructed. Beacon never computes a trend — the arithmetic happens first and Beacon phrases the result, so a fabricated improvement is not something a prompt could produce. And every reply is read in full before you see any of it, against a floor that judges what the reply claims rather than which words it contains; a reply that fails is withheld whole, because a partial refusal is a disclosure that has already happened.

It is not a clinician

Beacon provides health information and education only — not medical advice, diagnosis or treatment. It is never called Dr anything, it carries no medical badge, and no part of this platform gives individualised medical advice: not Beacon, not a verified clinician in a directory, not a table.

03 · Honestly

What is not built

Written down because a product page that lists only what works is a product page you cannot plan around.

  • Help with the cost of testing and treatment. The contribution rates are contractual and the mechanism is designed; the grant portal is not open.
  • Community forums and shared aggregates between patients.
  • Caregiver access. Delegated access needs its own design first — including how the record shows someone acting on your behalf rather than as you.
  • Matching you to clinical trials. The reference data is there; the match itself is an inference about your record, so it needs its own consent decision before it exists.

What is coming, in what order, and what has been dropped on purpose is on the public roadmap.

Where to go next