What changed
FACT (cited rule): FDA issued a June 2026 classification order placing radiological ML-based quantitative imaging software WITH a Predetermined Change Control Plan into class II with special controls, letting vendors ship model updates without a new 510(k). INFERENCE: the first PCCP-authorized updates are only now reaching hospitals, and no standardized evidence format exists for hospitals to verify an update stayed inside the cleared envelope.
Why now
The classification is weeks old; the evidence-format 'template window' is open before ACR/SIIM, CHAI, or incumbent credentialing players harden a standard. HYPOTHESIS: whoever defines the dossier format in the next 6-12 months becomes the default.
Converging signals
Single regulatory signal (the FDA classification rule) plus a pattern-instantiation hypothesis (Reverse-KYC Passport). This is one real signal and one inferred cascade — not a multi-signal convergence. The hospital re-review burden is plausible but UNPROVEN in the provided evidence.
Customer pain
HYPOTHESIS: hospital radiology QA / imaging-informatics / clinical-engineering committees have no defined process to validate a PCCP-based update and either freeze updates or accept them blind; vendors then field N bespoke questionnaires per release and lose deployments when hospitals freeze. No PAIN or HIRING evidence was supplied to confirm this — the demand_evidence array contains only the rule itself.
Who pays
Proposed payer: the AI-imaging vendor (per-release dossier fee), with hospitals subscribing per-seat. INFERENCE: vendors have regulatory budgets and lose revenue when updates stall. But note: the FDA rule compels vendors to satisfy FDA special controls, NOT to furnish hospitals a standardized dossier — so the 'forced buyer' is forced toward FDA, not toward this product. The passport purchase is voluntary on both sides.
Solved today
HYPOTHESIS: ad-hoc release notes, bespoke security/QA questionnaires (MDS2 for security), vendor customer portals, and in some hospitals nothing at all. ACR Assess-AI and CHAI assurance-lab/model-card efforts are adjacent emerging standards that could subsume this.
Why current solutions are bad
Many-to-many duplication: each vendor answers each hospital separately per release; no locked-test-set performance-delta format; hospitals lack PCCP literacy. All inference pending the proposed 10-interview test.
Proposed product
Per-release Update Passport: standardized dossier (release notes bound to PCCP envelope, performance deltas on locked test set, affected indications, rollback plan) published to a portal; hospital reviewers subscribe, receive, and attest. Vendor pays per release; hospitals per seat.
MVP version
A dossier template (structured PDF/JSON schema) + a minimal portal with vendor upload, hospital subscribe/attest, and an audit trail. Technically trivial for this founder; the hard part is not the software, it is convening both sides.
30-day build
Do NOT build. Run the stated falsification test: 10 interviews with hospital imaging-informatics/clinical-engineering leads (SIIM listserv, LinkedIn) — count how many have no defined PCCP-update validation process. Simultaneously scan SIIM/ACR forums for update-governance complaints and check whether ACR Assess-AI or vendor portals already standardize update evidence.
60-day build
Only if ≥4/10 confirm the gap AND no incumbent format exists: draft the dossier schema openly (spec-first, to become the default), and pitch 3-5 of the ~900 FDA-listed radiology-AI vendors on a paid pilot ($2-5k) for their next PCCP release.
90-day revenue plan
1-2 vendor pilots at $2-5k per release plus 2-3 hospital design partners (free initially). Realistic first meaningful revenue is 120-180 days, and only if hospital-side committees engage — the slowest actor sets the clock.
Distribution path
SIIM/ACR imaging-informatics community, RSNA vendor list, direct outreach to the FDA radiology-AI clearance list (public data — plays to founder's public-records strength). No paid acquisition needed.
Pricing hypothesis
Vendor: $1-3k per release passport or $10-20k/yr unlimited. Hospital reviewer seats: $50-150/seat/mo or free-to-read to solve the chicken-and-egg. All hypothesis — no willingness-to-pay evidence supplied.
Technical difficulty
Low-moderate: portal + schema + attestation workflow is a few weeks of AI-assisted build. The difficulty is standards adoption and healthcare credibility, not code.
Legal / regulatory risk
Moderate: the product summarizes regulated-device evidence; errors in a dossier that a hospital relied on could create liability. Founder has no medtech/regulatory credential, which matters in this buyer community. Not FDA-regulated itself (it is documentation tooling), but adjacent enough to demand care.
Platform dependency
None technically. Standards dependency instead: ACR, SIIM, CHAI, or a credentialing incumbent (symplr/GHX) publishing a competing format kills the wedge overnight.
Founder fit
POOR-TO-MODERATE, despite superficial resemblance to the ELDT win. The ELDT edge is: regulation forces party X to FILE INTO A GOVERNMENT PORTAL, founder automates the filing, charges per transaction. Here there is NO portal and NO compelled filing between the parties — the founder must invent a voluntary standard and convince two sides of a healthcare market (vendors' regulatory teams AND hospital committees) to adopt it. That is a network-effect, trust-cycle, quasi-enterprise healthcare play — three things this founder explicitly avoids. The applicable lesson ('government-portal mandate opportunities fit best', conf 0.80) argues AGAINST this idea, not for it: the mandate here runs vendor→FDA, not vendor→hospital.
Breakout potential
High if it worked: PCCP applies beyond radiology (cardiology, pathology, any adaptive SaMD), and the passport owner becomes the compliance rail for AI medical-device updates. But breakout requires becoming a de facto standard — a consortium-shaped outcome, not a solo-shaped one.
Final recommendation
PARK — do not build now. The regulatory window is real and recently opened, but the product is a voluntary two-sided standard in healthcare, not a compelled-filing automation, so it inverts rather than exploits the founder's proven edge. Spend at most 2 weeks on the free falsification test (interviews + ACR Assess-AI check); resurrect only if hospitals demonstrably block PCCP updates AND a vendor volunteers pilot money unprompted. Meanwhile, harvest the useful residue: the FDA PCCP clearance list is public data — a cheap 'PCCP update tracker' report/data product for hospital imaging-informatics teams would be a lower-commitment probe of the same demand.
Next action
Post one question to the SIIM community forum / listserv and message 10 hospital imaging-informatics leads on LinkedIn: 'When an FDA-cleared AI tool ships a PCCP-based model update, what is your process for validating it?' — target ≥4/10 saying 'we have none / we block updates' before spending another hour on this.