5 September 2026
In plain English
Quiet day on the numbers, loud day on the sourcing.
Two things are worth your morning. First, the US drug regulator spent this week telling reporters it is working with manufacturers to get more menopause oestrogen patches into pharmacies — and its own public shortage database, updated yesterday, does not list oestrogen patches at all. Both things are true. That gap is the most useful thing our menopause site can say this week, and nobody else is saying it.
Second, two business outlets reported that the drugmaker Novo Nordisk has settled or dropped its fight with the telehealth company Hims & Hers. We went looking for the paperwork and could not find it. So we did not move the forecast that depends on it, and we did not write it onto a page. That is the system working.
Signal detail
Historian: the reference class is first-year uptake of a new Medicare benefit, where 10–15% of eligible is typical; the present case is atypical because the Bridge sits outside Part D by design, so it does not inherit Part D's enrolment mechanics or its reporting cadence. Skeptic: the falsifier is straightforwardly the announced expansion — SurvivorNet, 2026-09-03, "Potentially Reaching Millions More Seniors" — and against an Oz-attributed ~600,000 at roughly two months in, a sub-1,000,000 December looks increasingly like the losing side of the bet. Bayesian: prior 0.40. Two new items pull opposite ways and the expansion is the louder and the newer, so a small move against: 0.38. Augur: nothing. The structural defect logged 2026-08-01 is unchanged and is now the dominant risk to this row — `cms.gov/newsroom` answered live again today at 200 and its six most recent releases are state infrastructure grants and a lab-fraud action, with no Bridge enrolment figure for a fourth consecutive run.
Historian: Case 2 in `historical-cases.md` is exact — lawsuit velocity from Novo and Lilly was the leading indicator of the compounded-semaglutide bust, so its reversal would be the leading indicator the other way, and it would be large. Skeptic: the counter-case got heavier this week and it is verifiable — an FTC action plus a wave of securities-fraud class actions (Law360 2026-09-03, Courthouse News 2026-09-02, Kaplan Fox deadline 2026-11-02). A company in that posture does not open a new compounded line on a discretionary timetable, and this one has retreated under enforcement before. Bayesian: the positive is unverified and the negative is real, which is an asymmetry, not a reason to move. No move, 0.60, and the check is named rather than deferred: EDGAR full-text returns four HIMS 8-Ks containing "Novo Nordisk", the newest dated 2026-03-11, and the browse-EDGAR feed shows no HIMS 8-K at all since 2026-08-10. A material definitive agreement would normally file Item 1.01 within four business days, so an absence on 2026-09-05 is not yet disproof — it is a clock still running to 2026-09-09. peptide-safety-incident: holds 0.95 (ceiling). Five more mid-tier safety items and no top-five outlet, which is now the seventh consecutive day of that exact shape. Augur, and this is the day's one weird thing: the FDA is reportedly screening GLP-1 ingredients at the border and warning about vials labelled with pharmacies that do not exist (Medical Daily, 2026-09-02). That is a new counterfeiting mechanic and, unlike "purity is bad", it is something a reader can actually check in ninety seconds against a state board licence lookup. It became a work order.
Falsifier queried explicitly and negative for a seventh run: the Katalys 7-day partner table (2026-08-28 → 2026-09-04) lists nine partners and zero peptide vendors, with Embody at $1,050 of the $1,400 payout. Clicks −15%, conversions flat at 11. Not quoting the payout decline as a market fact — the pulse's rolling window under-reports the tail.
Resolutions & upcoming
Nothing resolved. Nothing resolves inside 14 days. The next horizon is 2026-12-31 (three rows), then 2027-03-31.
How Barque got smarter today
- STEP V paid for itself, visibly. Two secondary items said Novo Nordisk had dropped its suit against Hims or struck a deal with it. Taken to primary, EDGAR carries neither. Had that gone into an order it would have been the 2026-08-02 defect verbatim — a named, honestly-cited source, and a wrong specific. `contradictions_found: 1`, `contradictions_fixed: 1`.
- The order-defect histogram is 44% illegible and I am reporting that first. 126 defects across 37 orders (118 corrected-and-shipped, 8 stopped-the-order; never summed). `UNCLASSIFIED` is 31 and `(not enumerated)` is 25, so 56 of 126 are classes I cannot read. The legible ranking underneath is OMISSION 18, PREMISE-FALSE 12, NUMBER 7. Two mechanical changes executed this run rather than intended: against OMISSION, the hrtpicks order carries its demand check inside the reason, run before the row was written; against PREMISE-FALSE, the glp1picks order names no route I have not read and instructs a grep of our own surfaces first. Against UNCLASSIFIED I can only propose, since I cannot edit my own brief — verbatim amendment in §6.
- Reroutes: found, and already re-filed. Two `REROUTE: BESTPEPTIDEFORTHAT` rows sit on killed glp1picks orders from 2026-08-03. Both were re-filed on 2026-08-02 and 2026-08-04 (`bestpeptideforthat-update-precision-peptide-legitscript-path`, `bestpeptideforthat-update-wapo-peptide-safety-investigation`). Not re-filing. Channel read, not skipped.
- An absence measured three ways is a finding; measured once it is a bug. openFDA returned NOT_FOUND for estradiol on `generic_name`, on free text, and on `openfda.generic_name`. Before treating that as real I asked the endpoint whether it was alive at all: 200, 1,634 records, last updated 2026-09-04. That third query is the one that turns a possible outage into a publishable fact, and it cost nothing.
- The demand answer was "no data", and it changed a decision honestly. DataForSEO sized the counterfeit-check cluster at `fake ozempic` 110/KD 5, with `counterfeit semaglutide` and `how to tell if peptide is real` not returned at all. That fails New Vertical threshold 2 on measurement, not on laziness — see §6. `estradiol patch shortage` returned no volume data, which the workflow itself labels restricted-or-no-data and NOT zero demand, so the hrtpicks order's `high` rests on the live news cycle and the verified primary contrast, and the row says so.
- Tier 2 is still dark and the counter is intact. `reddit_r_peptides` and `reddit_r_menopause` both 403 from a GitHub runner with a body naming the unblock. Logged under the URL-derived keys, not a tier name, because a key that moves cannot be counted.