idea-003 · narrative
A synthesis across this candidate's claims, dossier and decks — every statement here traces to something that already exists elsewhere in the knowledge base. A narrative never invents a fact, upgrades a tag, or recommends anything.
How to read this. This is a synthesis, not a new source of truth: every claim it cites carries the tag the candidate file gives it right now, copied not paraphrased, and every line quoted from a deck is quoted verbatim. See docs/narrative-spec.md.
This is a contrived plasma reference material — pooled human donor plasma spiked with characterized variant-bearing fragments at 0.001%-0.5% variant allele fraction, each level quantified by digital PCR — sold to molecular laboratories validating tumor-informed circulating tumor DNA (ctDNA) minimal residual disease (MRD) assays. It launches as a Research Use Only reagent, deliberately outside a patient-facing indication and, on the file's own framing, outside premarket review. Five claims are verified and every one of them is regulatory or definitional — two of them, read together, point at two different FDA regulations for the same material, and neither is confirmed to actually cover it. A business case was appended to the file most recently, adding five commercial assumptions — a year-three market-share target, a price premium over the incumbents, gross-margin sufficiency, a first-shipment timeline, and a competitor's buy-versus-build choice — every one of them recorded as unverified, meaning none has ever reached a Verifier pass at all. Everything downstream of the regulatory question — whether laboratories will buy it, whether the lowest allele-fraction levels can be made at all, and whether the customer's own reimbursed test volume is large enough to matter — rests on claims nothing in this repo has been able to check, most of them permanently so.
The product's launch claim is a labeling move: material shipped "For Research Use Only. Not for use in diagnostic procedures" under 21 CFR 809.10(c) is exempt from IVD labeling requirements, and the candidate's own frontmatter treats that as sitting the product outside premarket review entirely. That reading does not survive a read of the regulation's own text: 809.10(c) exempts a shipment from the labeling requirements in paragraphs (a) and (b) and from a part 861 standard, and "says nothing about premarket notification or approval" [claim 1: unconfirmed]. Whether the RUO label nonetheless holds up in practice turns on FDA's 2013 guidance on when RUO/IUO labeling is defeated by marketing conduct — a guidance document no connector in this repo can retrieve [claim 2: unconfirmed] — and on whether marketing this material specifically to CLIA laboratories validating clinical tests is exactly the conduct that guidance treats as disqualifying, the candidate's own stated principal regulatory risk [claim 3: unconfirmed].
A second, planned SKU repositions the same material as daily clinical QC. Here the file is better mapped: 21 CFR 862.1660 defines quality control material as Class I, verified against both the openFDA classification record and the eCFR text itself [claim 4: verified], and that classification is exempt from 510(k) premarket notification except for donor-screening use, subject to limitations in 21 CFR 862.9, also verified against primary text [claim 5: verified]. Product codes exist under that regulation [claim 6: verified]. But a Verifier-added finding complicates the frontmatter's own premise: FDA classifies a genetics/DNA quality control material as Class II under a different regulation, 21 CFR 866.5910, product code NZB [claim 22: verified] — and 862.9(c)(1) and 866.9(c)(1) carry the identical limitation, withdrawing the 510(k) exemption for an in vitro device intended for use in "the diagnosis, monitoring, or screening of neoplastic diseases," which an MRD-assay control sits adjacent to on its face. Which of the two regulations a contrived cfDNA MRD panel actually falls under is an open FDA intended-use judgment that neither eCFR text nor any connector here settles.
The demand case rests on 42 CFR 493.1256, CLIA's control-procedure standard, verified to require every molecular amplification procedure to run control materials at least once daily [claim 7: verified] — real, and mandatory, but the same verified text nowhere requires that material be purchased from a third party rather than built in-house, so it establishes recurring consumption without establishing that this candidate's product is what gets consumed. Whether laboratories currently find their own contrived controls inadequate is unconfirmed [claim 23: unconfirmed], as is whether they would prefer a third-party commutable panel over what they already build [claim 21: unconfirmed] and whether a CAP inspector or CLIA assessor would accept one as validation evidence at all [claim 24: unconfirmed]. The candidate names its own moat as the characterization dataset and the lot-to-lot continuity a laboratory anchors a published limit-of-detection claim to [claim 19: unconfirmed] — which requires laboratories to be willing to take on that dependency in the first place [claim 25: unconfirmed].
The reimbursement chain is entirely indirect and entirely blocked: the product itself carries no CPT/HCPCS code and none is achievable [claim 14: unconfirmed], so its market rides on the customer's own reimbursed test volume — CPT 0340U's existence [claim 8: unconfirmed], the Medicare payment rate for it [claim 9: unconfirmed], the annual Medicare claim volume behind it [claim 10: unconfirmed], and the MolDX coverage article naming it [claim 13: unconfirmed] — all unconfirmed for the same reason: the CMS cache files those checks depend on are absent from this checkout. The market-size components sit on softer ground still: QC spend as a share of lab test revenue [claim 11: unconfirmed] and the addressable segment of 30-60 laboratories [claim 12: unconfirmed] are both market-class claims this repo can never resolve, and whether FDA's 2024 laboratory-developed-test rule (vacated by a federal court, on the candidate's own account) leaves these assays under CLIA oversight is likewise unconfirmed and material to TAM in both directions [claim 15: unconfirmed].
On mechanism, the candidate names its own largest technical risk: material with a stated uncertainty budget at 0.001%-0.01% VAF is Poisson-limited at typical 10-30 ng cfDNA inputs, so the lowest panel levels may be irreducibly imprecise rather than merely hard to make [claim 16: unconfirmed]. The research lane has since added an empirical echo of that argument rather than a bench answer to it: a multi-manufacturer study across eight NGS platforms found that coefficient of variation rose as allele fraction fell, with poor repetition below 0.5% VAF [claim 39: unconfirmed], and a formal 89-laboratory external quality assurance program for cfDNA variant testing operated at 0.5%-2.5% VAF with over 92% concordance [claim 40: unconfirmed] — both real published findings sitting one to two-and-a-half orders of magnitude above the candidate's 0.001%-0.01% target floor, on domain-adjacent material rather than a bench test of this candidate's own proposed panel. Against that ceiling sit the published assays this material would validate: Haystack MRD's LOD of 0.063 mean ctDNA molecules/mL [claim 34: unconfirmed], CancerDetect's LOD of 0.001% VAF at 99.9% specificity [claim 35: unconfirmed], NeXT Personal's LOD95 of 3.45 parts per million [claim 36: unconfirmed], Invitae's Personalized Cancer Monitoring assay ranging 0.008%-0.05% VAF [claim 37: unconfirmed], and a research method (GeneBits/umiVar) benchmarked against commercial cfDNA reference standards reaching 0.0017% VAF [claim 38: unconfirmed] — a real and documented gap between what these assays claim to detect and what any named reference material is characterized down to, which is exactly the space this candidate is built to fill and, on the same evidence, exactly the space the discipline's own published reproducibility has not yet reached.
Competitively, the file names three vendors and no confirmed product characterized in the sub-0.01% VAF range on native fragmentomic patient matrix [claim 17: unconfirmed]. Two of the three have since been read: LGC/SeraCare's Seraseq ctDNA v4 line bottoms out at 0.1% VAF in a vendor-described "plasma like matrix" [claim 45: unconfirmed], and Horizon Discovery/Revvity's standards are fragmented human cell-line genomic DNA spiked into a synthetic plasma matrix, offered down to 0.1% [claim 46: unconfirmed] — neither reaching the candidate's target range, neither built on pooled donor plasma. The domain dossier reads this against the candidate's own mechanism claim directly: the published assay landscape spans roughly 0.000345% to 0.5%+ VAF across five distinct methods using units (molecules/mL, VAF percent, parts per million) that are not directly interconvertible without knowing input mass, and the top of the candidate's own stated target range, 0.5%, is "comfortably inside what commercial reference vendors already characterize" [dossier: ctdna-mrd-assay-validation]. Market growth for the downstream assays is reported by the vendors themselves — Natera's MRD test volume grew 34,000 units in Q2 2026 over Q1, its largest sequential increase to date [claim 43: unconfirmed], and Guardant Reveal grew more than 100% year-over-year for a third consecutive quarter [claim 44: unconfirmed] — both press and earnings-call sources no connector in this repo can independently verify. The dossier separately confirms it searched for, and could not find, any published record of a laboratory publicly reporting it switched away from an existing commercial reference material for a specific documented failure mode — the revealed-behavior twin to the purchasing-preference question the candidate itself flags as unanswerable, recorded here as not found rather than answered [dossier: ctdna-mrd-assay-validation].
A business case was appended to the file on 2026-09-03, adding five commercial assumptions the file itself records as never having reached a Verifier pass at all — the distinct meaning of unverified in this repo, as against unconfirmed for a claim the Verifier attempted and could not settle. By year three a third-party commercial reference panel is assumed to capture at least a low double-digit percentage share of addressable US MRD-laboratory reference-material spend rather than losing it to in-house contrived controls or the three incumbent vendors [claim 47: unverified]. A laboratory or assay manufacturer is assumed to pay a per-panel price materially above the incumbent vendors' list price for a panel characterized into the sub-0.01% VAF range [claim 48: unverified]. The resulting gross margin, once donor-plasma sourcing, orthogonal dPCR characterization, and ISO 17034-oriented documentation costs are included, is assumed high enough to support a standalone reference-material business at the addressable laboratory count and purchase cadence [claim 49: unverified]. First commercial shipment to a paying laboratory customer is assumed achievable within the same low-single- digit-million capital budget on a calendar of roughly 9-15 months from a standing start [claim 50: unverified]. And an existing reference-material vendor is assumed to prefer licensing or acquiring the sub-0.01%-VAF characterization and orthogonal dPCR quantification methodology over building it in-house, at a cost and calendar lower than fielding an independent direct-sales channel to the 30-60 laboratory segment [claim 51: unverified]. None of the five has been checked in either direction; they are recorded, in the file's own words, "so the Verifier can see them."
All three decks carry built: 2026-08-30, and python3 engine/scripts/check_decks.py --candidate idea-003 reports all three ok — none flagged STALE. That clean result is worth reading carefully rather than taken as evidence the decks reflect the file's current content: each deck's source_updated frontmatter field already reads 2026-09-03, matching the candidate's current updated:, but the five business-case claims (47-51) were appended to the candidate on 2026-09-03 — after all three decks were built on 2026-08-30 — and none of the three decks discusses them anywhere in their text. The checker verifies that the stamp matches, not that the deck's content was regenerated against it (docs/narrative-spec.md §8); here the stamp moved and the content did not.
The desirability deck turns on a single behavioral question the file itself names as an assumption and nothing has moved since: whether a laboratory will anchor its own published limit-of-detection claim to an external supplier's lot, which the deck reads as simultaneously the purchase decision, the recurring-revenue mechanism, and the candidate's own stated moat. Thirteen claims were added to the file since the deck's previous build — market growth figures, two incumbent vendor specifications, and five published assay LOD claims among them — and, on this deck's own accounting, none of them changed that this candidate's five verified claims are all regulatory or definitional and describe no customer at all.
"whether laboratories will anchor a published limit-of-detection claim to an external lot" — the load-bearing condition of the desirability deck (built 2026-08-30).
The viability deck turns on the customer's own economics — whether the downstream MRD test this product's revenue depends on is well paid and high volume — and names something the other two decks do not: the CMS cache files those three claims are blocked on were restored in this repository on 2026-08-30 and used to resolve claims on a different candidate the same day, but this candidate was not re-run against them. The deck reads that plainly as a scheduling gap rather than a research gap. Since this deck was built, the business case appended five more unverified commercial assumptions it never had a chance to weigh: a year-three market-share target [claim 47: unverified], a price premium over the incumbents' list price [claim 48: unverified], gross-margin sufficiency for a standalone business [claim 49: unverified], a 9-15 month first-shipment timeline [claim 50: unverified], and whether an incumbent vendor would license or acquire the method rather than compete with it directly [claim 51: unverified] — all five recorded as never having reached a Verifier pass, not as checked and found wanting.
"whether the customer's own test is well paid and high volume — the CLFS rate and the Medicare claim volume behind tumor-informed MRD testing" — the load-bearing condition of the viability deck (built 2026-08-30).
The feasibility deck turns on whether the lowest panel levels can carry a stated uncertainty budget at all, given the Poisson counting limit at realistic cfDNA input masses — the candidate's own named largest technical risk, previously argued from first principles alone and now sitting beside a published multi-manufacturer measurement of reproducibility breaking down below 0.5% VAF, roughly 50 to 500 times above the panel's target range.
"whether the lowest panel levels can carry a stated uncertainty budget at all, given the Poisson limit at realistic cfDNA inputs" — the load-bearing condition of the feasibility deck (built 2026-08-30).
Where the three agree is on what the newly appended research-lane and literature-detail claims did not do: all three decks independently note that thirteen new claims moved no tag, because vendor specification pages, press releases, earnings calls, and literature abstracts have no connector in this repo capable of resolving them to verified. Where they differ is in what each reads as the cheapest next step — the viability deck points at a data-restoration task that was already possible and simply was not run, while the desirability and feasibility decks both point at work (customer interviews, bench replicate studies) that has not yet started at all. None of the three had a chance to react to the business-case addition; that gap belongs to all three decks equally, even though the business case's substance — market share, pricing, margin, timeline — sits closest to the viability chair.
Given that the candidate's own launch regulation does not say what its frontmatter says it says, that a second verified regulation points at a different classification for the same material, that the discipline's own published reproducibility floor currently sits one to two-and-a-half orders of magnitude above the range this material would need to characterize, and that a business case has now staked out specific market-share, pricing, margin, and timeline assumptions [claim 47: unverified] through [claim 51: unverified] without any of them having reached a Verifier pass — does resolving the reimbursement chain behind the customer's test change anything about whether this product is buildable and sellable, or is the regulatory and technical footing underneath it the question that has to settle first, before the business case's own numbers are worth checking at all?