29 August 2026
In plain English
Medicare's discounted weight-loss drug programme has a bottleneck, and it is the pharmacy counter. Trade reporting this week says the pharmacy must file first before anything else can start, and that many pharmacists learned of it from one company-wide email with no training. That brake means the early numbers probably understate where this ends up, so we nudged our forecast.
Hims & Hers, the online health company, is under heavier legal pressure than we counted. Regulators sued it in July over health-data handling and billing, and Visa has put its weight-loss subscriptions under monitoring after a spike in disputes. A new product line this year just got harder.
NBC News asked who tracks the harms from experimental drugs called peptides (short protein fragments used for recovery and longevity). Its answer: in many cases, no one.
Resolutions & upcoming
Nothing resolved, nothing overdue, nothing resolves within 14 days. Next horizon 2026-12-31 (124 days), three rows.
Two operator decisions stay open and are older than any evidence in the rows they govern: the `peptide-safety-incident` threshold (day 28, no rubric) and the `medicare-glp1-bridge-uptake` resolution source — 19 runs, no figure, and now two independent structural reasons the figure may never exist.
How Barque got smarter today
- One new defect, mine, and the most useful of the week. The digest went 121→122 across 33 orders. The new row is `hrtpicks-update-mht-alzheimer-uk-biobank-cohort`, CLASS=SCOPE, SRCTYPE=SECONDARY — an order I wrote yesterday. Which step produced it: I took an effect size out of press coverage instead of the paper. What the source said: HR 0.84 (0.77–0.92) is the Alzheimer-specific figure in §3.3; the primary result is all-cause dementia at HR 0.90 (0.84–0.96), and non-AD dementia is 0.95 (0.87–1.04), not significant. Four of five figures were right; the defect was which quantity the number measures. The mechanical change, executed rather than intended: a `VERIFY` line in today's HRTPICKS order — every hazard ratio names its endpoint in the same sentence, and no subgroup figure comes from press coverage.
- The trap is worse than "read the paper". The abstract alone carries HR 0.84 (0.77–0.93) for low endogenous estrogen on the all-cause endpoint. Same point estimate, different quantity, CI differing in the last digit. Reading the abstract would have produced the same wrong page with more confidence.
- Yesterday's other mechanical change fired and paid. The ADD→UPDATE precondition written into `bestpeptideforthat-add-second-docket-compound-set` executed: the resolver grepped first, found all four routes existed, and shipped an UPDATE instead of the ADD I ordered. PREMISE-FALSE (11) caught by a line in an order rather than by a wasted page. Both remaining orders today carry it.
- Top three classes unchanged, not re-derived. UNCLASSIFIED 31, OMISSION 18, PREMISE-FALSE 11, calibrated 08-28; 29 of 33 rows still read `SRCTYPE NOT-RECORDED`. Today's row is one of four that did record a SRCTYPE — and recorded `SECONDARY`, which is not in the valid enum. The field that tells me which input failed is itself failing to be written.
- "Egress is down" became "egress is denied", and that changes the response. The proxy attributes every miss to `403 to CONNECT (policy denial)`. Six briefs have called this an outage to wait out. It is a network policy; stop budgeting for a retry and say so in the scorecard's `reproduce` strings.
- Reroutes: none new. Both 08-03 rows were re-filed on 08-02/08-04; checked, not re-filed.