idea-003 · viability deck
Every slide states a condition that would have to be true, then reports where it stands using the candidate file's own claim and its own confidence tag. A deck never upgrades a tag, invents a number, or recommends anything, and there is no ask slide.
The chair: the investor, and the operator who has to build a consumable business selling into a customer set small enough to name.
How to read this: every slide is a condition, not a conclusion. A bracketed claim reference points into knowledge-base/candidates/idea-003.md — the number is the claim's position in its ## Claims list, the tag is copied from it and never adjusted here. A [no claim] marker means nothing in the file speaks to the condition at all. No number appears here that is not in a claim, and there is no ask slide. See docs/deck-spec.md.
What changed since the 2026-08-24 build: this deck's cheapest listed task — read the incumbent vendors' specifications for their lowest characterised VAF and matrix type — was done, for two of the three vendors [claim 45: unconfirmed], [claim 46: unconfirmed], and the answer supports the competitive gap the candidate claims. The load-bearing condition, by contrast, is unchanged and now anomalous: see slide 7. Rebuilt from the file, not patched.
Would have to be true: A molecular laboratory would have to pay for this out of operating budget — funded, ultimately, by the reimbursement its own tests attract.
Where it stands: The product has no code and needs none; revenue depends entirely on laboratories' consumable budgets [claim 14: unconfirmed]. Whether that spend is run-rate or project-based is on record and unchecked [claim 27: unconfirmed], as is whether the buyer can purchase it without a capital process [claim 26: unconfirmed]. The customer's own revenue is better characterised than it was, though only through vendor reporting: Natera's MRD unit volume grew by 34,000 tests in Q2 2026 over Q1 [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 sources with no connector able to verify them.
What would settle it: Lab budget-holder interviews on where reference material sits — assay development capital or run-rate consumables.
If it's false: The addressable spend is a project budget rather than a run-rate, and the business is a series of one-time sales into a small account set.
Would have to be true: Not for this product — it has no code and needs none. What has to hold is that the customer's test is well reimbursed enough to fund a QC budget.
Where it stands: All three legs are unconfirmed for the same missing local cache, not for a negative answer: the CLFS rate [claim 9: unconfirmed] ("No local cache at data/hcpcs_payment_rates.csv"), the Medicare claim volume [claim 10: unconfirmed] ("No local cache at data/medicare_procedure_volumes.csv"), and the MolDX coverage article [claim 13: unconfirmed] ("No local cache at data/medicare_coverage_policies.csv"). The code descriptor itself is a fourth question of a different kind — CPT is AMA-licensed and cannot be reproduced here at all [claim 8: unconfirmed]. The regulatory environment the customer operates in has one moving part: the 2024 LDT final rule is claimed vacated in March 2025, leaving MRD assays as LDTs [claim 15: unconfirmed].
What would settle it: The CMS caches in data/ — see slide 7, because the status of that task has changed since the last build in a way that matters; the Medicare Coverage Database for the MolDX article; a licensed AMA source for the code descriptor.
If it's false: The customer's economics are the market's economics. If MRD testing is not well paid and growing, a QC panel for it is a small line in a shrinking budget. The vendor reporting on slide 1 argues it is growing; nothing here confirms it.
Would have to be true: A small customer set, spending a real fraction of test revenue on QC, at a price that makes the account count a business.
Where it stands: Four components, and the two largest are the cache-blocked ones: the CLFS rate [claim 9: unconfirmed] and the annual claim volume [claim 10: unconfirmed]. The QC spend ratio is the genuinely soft input the market connector reports as unverifiable [claim 11: unconfirmed], and the segment size is the Generator's own narrowing [claim 12: unconfirmed]. Whether price times account count supports a standalone business rather than a product line inside a larger reagent company is this deck's own write-back [claim 29: unconfirmed] — and the two vendor specifications now in the file [claim 45: unconfirmed], [claim 46: unconfirmed] identify exactly which larger reagent companies that comparison is against.
What would settle it: Restore the CMS payment and procedure-volume caches and re-run; enumerate the laboratories from the MRD assay literature and MolDX-covered test lists; then the QC ratio is the only genuinely soft input left, and it can be replaced by direct budget questions.
If it's false: TAM is a hard kill condition, and a small-account market with a low ceiling would fire it on evidence rather than on absence. It has not fired, because the evidence has never been retrieved.
Would have to be true: A small wet lab, a digital PCR platform, donor plasma sourcing and ISO 17034-oriented documentation would have to be the entire bill — no instrument development, no clinical trial, no premarket submission.
Where it stands: The capital claim is unbenchmarked [claim 20: unconfirmed], and the "no premarket submission" premise resting under it is the strongest verified material in the file: Class I under 21 CFR 862.1660 [claim 4: verified], exempt from 510(k) except for named uses this product does not claim [claim 5: verified], with product codes existing under that regulation [claim 6: verified]. Set against it, a Verifier-added finding that FDA classifies a genetics/DNA quality control material as Class I under a different regulation [claim 22: verified] — two verified claims pointing at two regulations for the same material. The accreditation line item is unchecked [claim 33: unconfirmed].
What would settle it: A regulatory-counsel read on which regulation this material falls under given its intended use, and whether the RUO label makes the question moot or merely deferred; an ISO 17034 accreditation cost and timeline.
If it's false: The "no premarket submission" premise is doing the work in this factor, and a Class II designation would change cost and calendar materially — though only if the RUO position fails, which is the feasibility deck's problem.
Would have to be true: The characterisation dataset and the lot-continuity relationship would have to be defensible — because the candidate concedes the material itself is reproducible by any competent laboratory.
Where it stands: The moat claim rests on the characterisation dataset and the customer's willingness to anchor to it [claim 19: unconfirmed], which is the same thing the desirability deck names as its load-bearing condition [claim 25: unconfirmed]. Whether CAP assessors credit third-party documentation is unbacked [claim 24: unconfirmed]. FTO — the only structural barrier available on a product conceded to be reproducible — was never run and names no patent number [claim 18: unconfirmed]. The competitive picture is sharper than it was: two of the three named vendors [claim 17: unconfirmed] now have their product lines specified in the file, both bottoming out at 0.1% VAF in non-native matrices [claim 45: unconfirmed], [claim 46: unconfirmed].
What would settle it: Whether CAP assessors credit third-party documentation; whether the existing vendors' customers exhibit lot lock-in; a patent landscape on fragmentomic-profile and variant-spiking methods.
If it's false: A reproducible consumable sold to a customer set that can build it — with established vendors already in the category [claim 17: unconfirmed].
Would have to be true: The sub-0.01% VAF, native-fragmentomic-matrix niche would have to be genuinely unoccupied — and to stay unoccupied long enough to matter.
Where it stands: Two-thirds answered since the last build, in the candidate's favour. The last deck named a catalogue read as the cheapest competitive fact available; it was done for two of the three vendors, and both stop where the candidate said they would. LGC/SeraCare's Seraseq v4 line bottoms out at 0.1% VAF in a vendor-described "plasma like matrix" [claim 45: unconfirmed]; Horizon/Revvity's standards are fragmented cell-line genomic DNA at an average 160 bp in synthetic plasma matrix, offered down to 0.1% [claim 46: unconfirmed]. Neither reaches the sub-0.01% range and neither uses pooled human donor plasma, which is the gap the frontmatter claims [claim 17: unconfirmed]. Twist Bioscience, the third named vendor, has not been read — though a published method benchmarked itself on three commercial standards including a Twist product [claim 38: unconfirmed]. Both vendor claims are unconfirmed because no connector in this repo verifies a vendor specification page.
What would settle it: Twist's specification, to complete the set. Then a standing watch on all three, because the finding that matters is not the current gap but how long it survives.
If it's false: Read it inverted — the risk is not that a competitor appears, it is that the gap is a product-line extension for a company that already has the customers, the accreditation and the sales force. Two of the three now have documented floors one order of magnitude above this candidate's target, which is a shorter distance for them to travel than for a new entrant to travel the whole way.
Would have to be true: Three CMS cache files and one more vendor spec sheet would have to settle the money question.
Where it stands: This slide has changed in a way worth reading carefully. Three claims here are unconfirmed for a missing local cache rather than a negative answer [claim 9: unconfirmed], [claim 10: unconfirmed], [claim 13: unconfirmed]. Those caches were populated on 2026-08-30 — idea-001 in this same portfolio records all five cache files present and re-verified nine claims against them that day. This candidate was never re-run. So its three cache-blocked claims are not waiting on a download that has not happened; they are waiting on a Verifier pass that was never scheduled, against files that already existed once. The vendor-specification task is two-thirds done [claim 45: unconfirmed], [claim 46: unconfirmed].
What would settle it: Restore the CMS caches per data/README.md — they are gitignored, so they do not travel between sessions — and re-run payment 0340U, procedures 0340U and coverage 0340U. Three claims, one pass, no new research. Then Twist's spec sheet, then the laboratory calls.
If it's false: Nothing here is false — this is a work order, and it is shorter than it was.
Every condition above with nothing verified behind it.
One thing to read carefully: this file's only verified claims bearing on viability are [claim 5: verified] (exempt from 510(k) under 862.1660) and [claim 22: verified] (FDA classifies a genetics/DNA quality control material as Class I). Both are true and they describe different regulatory homes for the same material — which is precisely why the regulatory position, not the market, is the first thing to settle.
If only one thing from this chair could be checked: 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 [claim 9: unconfirmed], [claim 10: unconfirmed].
This product has no reimbursement of its own by construction [claim 14: unconfirmed], so its entire market is a fraction of a fraction of someone else's revenue: test volume, times payment per test, times the QC spend ratio [claim 11: unconfirmed], times the share of labs that buy externally [claim 21: unconfirmed]. The first two terms are the largest.
It stays the load-bearing condition, and it has become the most frustrating one in the portfolio. The last build described these two as "sitting behind a missing CSV in data/". The CSVs were obtained on 2026-08-30 and used to resolve nine claims on another candidate the same day. This one was not re-run. The cheapest, highest-yield work on this candidate is therefore not research at all — it is scheduling a pass that was already possible.
Naming it is not a recommendation, a gate, or a kill.