The complete index — every link verified live
Everything HashCare, on one page.
The proof layer for clinical AI: free anchoring, verifiable receipts, a published review standard, and a registry countersigned into Bitcoin every day. This page is the whole map — pages, proofs, endpoints, stories, and the primary sources underneath them.
The product surfaces
- hashcare.comthe front door — the July-2026 moment (four things now true at once) and the live tamper test: break a sealed record, watch the fingerprint catch it
- /anchorfree proof-of-existence: any file or text, hashed in your browser, never uploaded; public registry with moderated labels
- /verifypaste any SHA-256 — checks samples, clinical registry, care-hour receipts, and the anchor registry; found anchors show their Bitcoin countersign
- /real-reviewThe Five Tests of a Real Review v1.0 — the determination-level standard for verifiable human review; hash-anchored in its own registry
- /o1receipts for Medicare's proposed SaMS "O1" AI payment category — the per-use proof CMS's own rule says is missing
- /ledgerthe Judgment Ledger — a physician's accumulated receipts as a portable, independently verifiable career record of how they judge; owned by the reviewer, provable without us
- /homereceipts for the sovereign home — copy-paste Home Assistant config: one salted daily fingerprint per home, hashed locally, data never leaves the house
- /sandboxsubmit artifacts and generate demo attestations (demo personas, no real data)
- /swipethe physician review gesture, demonstrated
- /legalterms, privacy, health-data posture — no sale of data, no training on user data without consent
The documents & proof files
- real-review-v1.txtcanonical text of the standard · SHA-256
d731ce0ad779ec950a9167e91cf1e7d1b2afd84d7afb5af73874bd03d8bf4d38 · verify it
- roots/latest.jsontoday's registry root — anchor count, root hash, method; a new root is stamped into Bitcoin daily via OpenTimestamps
- roots/<date>.otsthe daily Bitcoin timestamp proofs (from 2026-07-14 onward) — verifiable at opentimestamps.org without trusting HashCare; older roots carry final Bitcoin block attestations
- llms.txtthe machine-readable version of this page — how AI agents verify, anchor, and cite HashCare
- og.jpgthe share image — "AI drafts it. Who signs it?"
The API — auth-free, works today
- Verify a hash
POST …/rest/v1/rpc/anchor_verify {"p_hash":"<64-hex>"} → found, label, anchored_at
- Anchor a fingerprint
POST …/rest/v1/anchored_artifacts {"hash","label"} — one call per receipt; keys on the page
- The O1 receipt recipehash of input + output + algorithm version + NPI, anchored per billed use — the full pattern with example calls
The stories — each one paragraph, each load-bearing
Four things are now true at once: TEFCA moved a billion records with DOJ-enforced pipes; CMS's WISeR model made licensed-clinician review of AI decisions a payment requirement; ChatGPT for Clinicians made AI drafting free for every licensed clinician in America; and CMS proposed paying for clinical AI per use. Free drafting, mandated review, enforced pipes, per-use payment — and still no proof of what was done, by whom, under whose license. That question is the entire product.
When China opened the world's first AI-native hospital in March 2026, it published the consensus definition of an AI hospital the same day. In the US, Joint Commission, URAC, and NCQA certify organizations, and WISeR requires review per determination — but nobody had defined what counts as a real human review of a single AI decision. The Five Tests fill that gap: licensed and matched, independent, engaged (measured), receipted, durable. Published free, versioned, and hash-anchored in the registry it prescribes — so priority is provable forever.
The CY2027 outpatient rule creates the first Medicare payment category for clinical AI: 36 codes, up to $950.50 per analysis, paid to the facility. In the same section, CMS admits per-click billing "raises concerns regarding program integrity." A four-field receipt — content hash, algorithm version, anchored timestamp, reviewer NPI — closes the gap for effectively nothing, and it's generatable today. The docket that decides the category's rules is open until August 31, 2026.
Every day, the whole registry is rolled into one root hash and stamped into Bitcoin via OpenTimestamps. Anyone can recompute the root from the public hash list and check the timestamp against the Bitcoin blockchain — no trust in HashCare required. If HashCare vanished tomorrow, every receipt would still verify. A proof that dies with its vendor was never a proof.
The NOHARM benchmark (Stanford/Harvard, 57 authors) found AI-assisted physicians scored lower than AI alone as practiced — presence isn't engagement. Bob Wachter, chair of medicine at UCSF: "You can't train someone to pay attention on the 40th time if it's been right 39 times in a row." That's why Test 3 requires a measured disagreement rate, not an assumed one — and why this site claims accountability, never accuracy.
For WISeR vendors and coverage-decision AI: state-matched, NPI-verified physician review delivered as an API, a receipt per determination, double-review calibration statistics, and a fixed-price two-week pilot on 50 real determinations. Scored against the published Five Tests, test by test. The network is ClinicalSwipe; the receipts are HashCare; the pipe is the harness.
Anchoring costs nothing and always will — proof of existence is infrastructure, not a toll. What's priced is judgment: licensed physicians reviewing clinical AI output, with receipts attached. The rail stays free so the receipts become universal; the signature is the product.
Primary sources (what everything above cites)
- CMS-1850-PCY2027 OPPS/ASC proposed rule — the SaMS/O1 category, §X.B (Federal Register, July 7, 2026)
- Docket CMS-2026-2344where comments on the O1 category are filed — open to everyone until Aug 31, 2026
- CMS WISeR modelthe rule that made licensed-clinician review of AI prior-auth decisions a payment requirement (six vendors, six states, 2026–2031)
- NOHARMthe Stanford/Harvard medical-safety benchmark and physician-teaming RCT behind Test 3
- CMS NPPESthe public NPI registry every reviewer credential is checked against
- OpenTimestampsthe open Bitcoin timestamping standard the daily countersign uses
- Joint Commission + CHAIthe org-level AI certification — the layer above the determination, where existing standards stop
The network
What's deliberately not here: user content (never uploaded — only fingerprints exist), pricing theater (anchoring is free, forever), and claims that human review improves accuracy (the evidence is mixed; this whole stack claims accountability and verifiability, which are provable). Every link on this page returned HTTP 200 on the day it shipped; the registry root above re-proves itself daily.
