11 October 2026
In plain English
Several sites reported this week that the US government had moved to allow the sale of experimental drugs called peptides — short protein fragments sold for recovery and longevity. Barque checked the Federal Register, the official record of US rule changes, twice. Nothing has been filed. The stories re-tell a July vote by the FDA's peptide advisory panel, which recommended loosening the rules. The government has not acted on that recommendation. Making these drugs at a specialty pharmacy is still not allowed today.
Separately, a Swiss hospital published the case of a woman in her twenties who spent five days vomiting after four weekly injections of an unverified product, given to her by a life coach with no medical training. She did not think of the injections as medicine.
Quiet elsewhere.
Signal detail
Primary: sec.gov browse-EDGAR (HTTP 200, 18,838 bytes) shows one Hims & Hers 8-K since 2026-08-10 — accession 0001773751-26-000220, filed 10-07, Item 5.02, an officer appointment. No Item 1.01 since 07-01, so the four-business-day clock this log set on 09-06 for the reported Novo Nordisk agreement expired 09-09 and is 32 days stale. Eighty-one days remain, and the "eyes peptides expansion" line dated 10-09 traces to that same Q2 call. What keeps this near even money: GHK-Cu is inside the forecast's list and needs no bulks listing, and a launch needs no filing. peptide-affiliate-share: 0.90 → 0.85. FLAGGED — Skeptic and Bayesian diverge by 0.15. After nine runs recording this falsifier as negative, it fired. The Katalys table for 10-03 → 10-10 carries Collective Peptides at 47 clicks, 1 conversion, $150, EPC $3.19, and $150 / EPC $3.95 in the 10-06 pack; on 09-07 the same table held nine partners and zero peptide vendors. Bayesian: 0.85 — $150 of $1,550 is 9.7% against a 30% threshold, and resolution reads a full-year file in which most months carry no peptide revenue. Skeptic: 0.70, on a new argument — this is a mix forecast and the denominator is moving. Embody alone is 68% of payout, two sister properties are deindexed, and the mix crosses 30% with peptides flat if weekly payout falls to roughly $500. Both logged. Against myself: collectivepeptides.com returned 200 at 162 bytes, empty body, so whether this partner is a peptide vendor or a GLP-1 brand using the word is unverified.
CMS answered live (HTTP 200, 46,756 bytes) and published nothing on enrolment. New is mechanism, not volume: a Center for Medicare Advocacy denied-coverage appeals explainer on 10-08, and the Boston Herald on 10-07. An advocacy body writing an appeals guide is evidence denials are at volume, and the Historian's class is that first-year uptake of optional, application-gated Medicare benefits lands below announced eligibility, every time. peptide-safety-incident: holds 0.95, where it cannot rise. New primary harm literature: BMJ Case Reports 2026 Oct 7;19(10):e276106 (PMID 42843917) — a woman in her mid-20s, five days of refractory nausea after four weekly dose-escalated injections of an unverified product reported to contain retatrutide, given by a self-described life coach with no medical qualification, Naranjo 7. The finding that travels: she did not consider the injections to be medication, so the exposure evaded routine history-taking. Premortem (Sunday, rotated to `peptide-safety-incident` — closest to a resolution judgement, and at ceiling where being wrong costs most). Suppose this resolves FALSE. The story never gets its top-five feature; it is absorbed as a liability beat. The latent condition is visible today: eleven weeks of trade and syndicated coverage, a case report confined to a specialist journal, and the one top-five item framing harm as litigation over branded weight-loss drugs rather than gray-market peptides. A saturated warning cycle inoculates. Implication: 0.95 is a ceiling, not a measurement, and the missing rubric is the real exposure. Augur. The Precision Peptide Company ordered 20,000 Thymosin Alpha-1 transdermal patches for commercial launch (trade press, 10-07) — consumer-goods scale, not compounding scale, and a patch sidesteps the needle objection. New forecast candidate (operator approval; Ra never auto-adds). FDA does NOT publish a proposed rule adding any PCAC-recommended peptide to the 503A bulk drug substances list in the Federal Register before 2027-03-31. Probability 0.70, peptides-longevity, 171 days, signal strength 63. Falsifier: any rule naming BPC-157, KPV, TB-500, MOTS-c, Epitalon or Semax. Resolution source: the Federal Register API, which publishes this object at a known cadence — chosen against the resolution-source defect open on `medicare-glp1-bridge-uptake` since 2026-08-01.
Resolutions & upcoming
Nothing resolved. Four pending, zero overdue, and nothing resolves inside 14 days: three land 2026-12-31 (81 days out), one 2027-03-31. Resolved track record unchanged at three — Brier 0.0625, 0.4900, 0.2025, mean 0.2517; `peptides-longevity` sits at 0.3463, above the 0.25 drift line.
One forecast needs an operator decision, not a probability. NPR published "Ongoing lawsuits allege GLP-1s can cause serious harm. But we don't know how often" on 2026-10-10. NPR is one of the five outlets named in `peptide-safety-incident`. Whether it resolves the row turns on two questions the row does not settle: whether GLP-1 receptor agonists count as "peptides" (semaglutide is one), and whether the lawsuits concern compounded or branded product. That is the missing-rubric defect logged 2026-08-01, now 71 days open and load-bearing. I have not auto-resolved and will not.
How Barque got smarter today
- The honest headline is that Barque was not running. `health/2026-09-08.json` through `2026-10-10.json` grade this routine RED on 33 consecutive days — "routine was due (4 4 *) and emitted no scorecard" — and no `health/runs/` directory exists after 2026-09-07. Not only Ra — `health/2026-10-09.json` records 4 of 4 due routines RED, the three product resolvers alongside this one, while the Actions layer committed digests, GSC snapshots and these health files daily. The alert existed and was right for a month, into a file nobody opens. A grade is not an escalation. Filed as `barque-fix-routine-fleet-dark-no-escalation`.
- My highest-density input lied to me, and the file itself said how to catch it. The pack reports glp1picks.com impressions "34528 ↓ -87%". False: 34,528 is the exact sum of four days (10-03 to 10-06), because Search Console lags, while the prior window held a real spike peaking 63,241 on 09-28. On the daily baseline, clicks rose from 25–44/day to 44–93/day and position improved 11.2 → 6.9. The `daily_90d` block carries its own warning — "a window that loses a past spike reports a collapse that did not happen" — and the aggregate above it does exactly that.
- Two sister sites died in the same week, and three open orders were treating it as three problems. hrtpicks.com went 1,795 → 602 → 35 impressions/week; bestpeptideforthat.com went 735 → 337 → 11. Same cliff, same week beginning 2026-07-26, both flat at single digits for ten weeks, while glp1picks and glp1pets grew straight through it. Two of five properties, both playbook forks, is a shared-cause signature, and a cross-property read no single product routine can see — which is the one thing this layer is for. Cause unverified; the orders say read robots, meta robots, canonical host and sitemap first.
- Both return channels answered, ordinarily. No order defects in the last 14 days — fetched and empty, not unread; widened to 40 days it returns 2, both CORRECTED AND SHIPPED, classes NUMBER and PREMISE-FALSE, both predating the dark period. The two REROUTE rows in `work-order-status.tsv` were both re-filed to BESTPEPTIDEFORTHAT already, so no reroutes this cycle.
- Brief amendment I cannot make myself, verbatim for the operator. In STEP V under "Routing, measured and not negotiable", after the fda.gov sentence, add: "hhs.gov blocks GitHub runners the same way, measured 2026-10-11 in run 38110849192: HTTP 403 at 398 bytes with an Akamai `errors.edgesuite.net` body, and `looks_bot_blocked()` tests only for `abuse-detection` and `apology_objects`, so the Archive fallback does not fire. Route HHS claims to federalregister.gov, and fetch PubMed through `eutils.ncbi.nlm.nih.gov/.../efetch.fcgi`, not `pubmed.ncbi.nlm.nih.gov/<pmid>/`, which returns HTTP 203 'Cookies must be enabled'."
- One schema defect logged, not fixed: `work-orders.tsv` rows 13 and 14 carry ten columns against nine; append-only discipline says I do not edit prior rows.
- Voice calibration. The Discourse voice earned its place twice, catching that both the "HHS legalizes peptides" wave and the "Hims eyes peptides" line are re-syndications of July and August with nothing new behind them. But the Skeptic is what changed an output: it found the denominator argument on `peptide-affiliate-share` that nine prior runs missed while they confirmed the numerator. Nine consecutive "falsifier queried, negative" entries were a monitor pointed at one half of a ratio.