Documentation
Everything, in one place, with the page count on it.
Every document this project has produced, the method first. Nothing here requires a key, a form or an account. Each entry says who it is for. Page counts are read from the files themselves rather than typed, so they cannot drift out of date. What changed, and when.
White paperThe Plumbing Was Never the PointInteroperability, data, and the unclaimed value in critical-care therapy management. Free.Read itStart hereThe method, in plain languageOne patient, one night: what the machine reads, what it says, what it cannot check.Read itDatasheetAI MedAgent technical datasheetContinuous medication reasoning over live clinical data, and the integrity checks first.Open itDatasheetEHR Integration datasheetFHIR R4 and CDS Hooks, what each vendor sandbox returns, and what it refuses.Open itDatasheetCDS Hooks datasheetThe eight medication screening classes and the coverage figure on every card.Open itArgumentFive Minutes and a ReasonWhy a five-minute summary of device data, and what reasoning does with it.Read itSpecificationPCD-E Sandbox Ingest SpecificationHow device data is accepted by the sandbox endpoint. PDF.Open the PDFCriteriaInfusion Pumps on Hospital NetworksEvaluation and wireless test criteria with scoring, vendor neutral. PDF.Open the PDFRecordWhat changed, and whenEvery change to the site and its documents, newest first, corrections included.See the log
The bill, the medication pass, and Ami
Three newer pieces. The bill is policy, published on the author's own site. Med Management is the whole hospital medication process in one interactive page. Ami is a consumer experiment that keeps everything on the phone.
A citizen's model bill on the Meaningful Use model: the eight-stage AI adoption ladder in statute, independent certification, incentives, oversight, and enforceable clinician protection. Rev 3, 27 October 2026.
Web page, danielpettus.com · For: congressional staff, policy groups and health system leaders
The same draft, formatted to print or forward.
PDF, 17 pages, danielpettus.com · For: same, on paper
One synthetic patient through every hand-off of the hospital medication process, and where a reasoning layer fits at each step. Interactive; a concept, not a product.
Web page · For: pharmacy, nursing and IT leaders
What the iPhone app collects, what leaves the phone, and what TestFlight testing adds. Short version: nothing about you leaves your phone.
Web page · For: anyone testing Ami
Standing Watch, ICU to home
The design pitch is the design document. Each demonstration in the Standing Watch section maps to a slide in it.
Between the rooms, nobody is watching. Three earlier tries, why now, the method, one watch across five rooms, the bridge to home, what exists, where the money is, and the limits. Live links to the demonstrations.
PDF, 13 pages · For: anyone deciding whether this is worth a conversation
The four steps, the five rooms, the mediator field by field in FHIR R4, the SMART Health Link bridge, what runs today, and the limits.
PDF, 3 pages · For: engineers, informaticists and device makers
The same sheet in a browser.
Web page · For: same, in a browser
The reasoning method
This is the point of the site. Everything else here exists to show the method survives contact with a real record and a real device. Start with the plain-language walkthrough if you do not work in health IT.
What the machine reads, what it says, what it could not check, and what it costs, with no acronyms. Written for the reader of the book rather than the reader of the standards.
Web page · For: anyone
The method itself: reading the trajectory between the moments a human is scheduled to look, using a large language model and public federal drug data.
PDF, 2 pages · For: engineers and informaticists
The FSD controller and what it demonstrates.
PDF, 2 pages · For: anyone who has heard "closed loop" and wants to know what is different
The method in a browser.
Web page · For: engineers and informaticists
The controller in a browser.
Web page · For: same, in a browser
EHR integration, and what the vendors actually returned
Published capability statements say what a server supports. They do not say what it will give you. Those are different questions and only one can be answered by reading documentation.
Epic, Oracle Health, MEDITECH and athenahealth compared on what their sandboxes actually served, with every request and response. Includes the resources Epic refused after granting the scope.
PDF, 9 pages · For: interface engineers, informaticists, CIOs
The same findings without the protocol detail, for a reader who needs the conclusion rather than the packet capture.
PDF, 6 pages · For: anyone
The interactive record of the integration work, vendor by vendor.
Web page · For: interface engineers
The datasheet in a browser, including the returns table.
Web page · For: interface engineers
How AI MedAgent runs as a CDS Hooks service, which is the mechanism that lets an outside reasoning engine land inside the EHR at the moment of the decision without replacing anything.
PDF, 5 pages · For: EHR analysts and clinical informaticists
The same, in a browser.
Web page · For: same, in a browser
Medical device interoperability
Why the record of what a device did is unreconstructable, and the endpoint built to test whether anyone will fix it.
The white paper. West Health measured the value of medical device interoperability in 2013 and put it above $30 billion a year. This argues the plumbing was necessary and was never the point: the value is in what flows through it, joined to the public knowledge assets and reasoned over at the bedside. Two ordinary ICU cases, current state and future state. Every number cited to its primary source.
PDF, 17 pages · September 2026 · For: CMIOs, pharmacy and nursing informatics, pump and EHR vendors, IHE and HL7 people
The memorandum. Why an infusion record cannot be reconstructed from what the standards actually carry, what the economic case looked like when it was made in 2013 and why it failed, and the seven principles behind the sandbox. Thirty-seven cited sources.
PDF, 20 pages · For: device manufacturers, IHE and HL7 people, biomedical engineers
How to connect. Two doors, the epoch schema, the conformance block the endpoint returns, the terminology mapped to IHE DEV TF-3 sections 7.1 and 7.2, the synthetic-data rules, and a sixty-second quickstart. Built entirely on published standards.
PDF, 14 pages · dated 31 August 2026 · For: device manufacturers and integration engineers
New in 0.4. Section 10a documents the conformance block every response carries: which of the nine IPEC events you actually sent, how your pump status and mode values trace back through the revisions, whether your MSH-21 assertion agrees with your trigger event, and how your delivery events pair up. Section 8.3 covers the two incompatible ways KVO is encoded in the field. Both sections say plainly that the pair readings are this sandbox’s inference and not a requirement of the standard.
GET /device still returns the live behavior of the endpoint and needs no key.
The same case as the memorandum, readable in a browser, with the figures and the live connection instructions.
Web page · For: same, in a browser
One file, zero dependencies, Node 18 or newer. Mints its own key, generates a synthetic five-minute pump window, sends it and prints what came back. Published so that "the endpoint was broken" is never an available answer.
JavaScript module · For: developers
What the sandbox is, the two doors, the rules, and why a manufacturer would bother.
Web page · For: device manufacturers
Live endpoints
These are running now and answer without a key. They are generated by the deployed code, so they always describe what is actually running rather than what the documentation claims.
facts.json The vendor findings as structured data, which is the same file the datasheet and the evidence page both read from, so the three cannot disagree.