Barque · The Compound Research·Vol. 01 — Est. 2026

The ship we send
into the signal,
returns at dawn.

Barque is the research engine behind every Compound decision. One operator, a dataset far larger than any one mind can hold, and the discipline of curiosity applied to weak signals most desks ignore. The protocol fuses them across domains no single team bridges and returns falsifiable, Brier-scored forecasts.

¹Named for the solar barque that carried Ra through the Duat each night — the ship returns at dawn with what it learned in the underworld.

Scroll · The log below

Pending forecastResolved true

IILive forecasts

Updated 15 September 2026

Predictions with resolution dates.

Each forecast is a specific, falsifiable claim with a probability and a date it resolves. Probabilities are floored at 0.05 and capped at 0.95 — never absolute, never hedged. When a forecast resolves we score it with Brier and publish the result, even when it hurts.

GLP-1 / Metabolic

Medicare GLP-1 Bridge

CMS reports fewer than 1,000,000 Medicare GLP-1 Bridge enrollees by 2026-12-31 (per official CMS Part D enrollment data)

Base rate: new Medicare Part D programs enroll slowly in year one (10-15% of eligible). 1M = ~7% of 14M eligible. Launches July 1 2026, so only 6 months of enrollment window in 2026. Friction factors: physician re-prescribing, formulary navigation, step therapy. Counter-narrative ~35%: aggressive CMS outreach + pent-up demand drives faster-than-normal uptake. Resolves on CMS public enrollment data release (usually Q1 following year).

peptide-affiliate-share

By 2026-12-31, peptide vendor affiliate revenue is less than 30 percent of total The Compound affiliate revenue (i.e., GLP-1 telehealth remains dominant revenue stream)

Math dominates: peptide affiliate CPA is 10-30x lower than GLP-1 affiliate CPA. Even if peptide traffic matches GLP-1 traffic, revenue share remains minority. Counter-narrative ~15%: PCAC grants Cat 1 (30% chance per prior forecast) + major US pharma launches approved peptide product with affiliate program + Compound pivots heavily. Resolves on The Compound internal Katalys + peptide network payout data. | OPERATOR 2026-04-23: Foundayo FDA-approved 2026-04-01; Hims listing it at $25/mo insurance / $149 cash — branded-GLP-1 economics more entrenched, marginally strengthens 0.85.

By 2026-12-31, Hims and Hers publicly launches a compounded peptide DTC offering (BPC-157, GHK-Cu, or TB-500) through its telehealth platform

Hims acquired the facility 14 months ago for exactly this. Business-line extensions typically launch within 18-24 months of M&A. Reclassification gives political cover. Counter-narrative ~20%: regulatory setback at July PCAC forces Hims to delay to 2027; safety incident in industry triggers pre-launch freeze. | OPERATOR 2026-04-23: Foundayo (Lilly orforglipron, FDA-approved 2026-04-01) now distributed via Hims at $25/mo insurance / $149 cash alongside Wegovy pill — confirms Hims-as-channel-for-every-approved-GLP-1 thesis. Active behavioral pattern: in Feb 2026 Hims sold a compounded copy of the first GLP-1 pill 2 months post-approval; FDA forced retreat within weeks. Treat as reinforcing 0.83 with retreat-under-enforcement as a latent cap on unrestricted launch timing.

peptide-safety-incident

By 2027-03-31, a top-5 US news outlet (NYT, WSJ, CNN, NPR, WaPo) publishes a major feature on peptide injuries, contamination, or deaths sourced from compounded/gray-market products

Adoption at current velocity plus 40 percent contamination rate plus minimal regulation equals statistical certainty of adverse event. Media narrative arc typically runs hype-phase to backlash-phase within 12-18 months. Already have safety-focused coverage seeding (NPR Feb 2026, UNSW April 2026). Counter-narrative ~30%: FDA enforcement of gray-market sellers tightens fast enough to prevent headline-grade incident; industry self-regulation via compounders catches contamination early.

VThe Dawn Log

The ship returns. We log what it saw.

Ra runs each dawn. Every brief opens in plain English before the rigorous layer — forecasts, Augur picks, resolutions, and the meta-cognition of how Barque got smarter. On Sundays it expands with a rotated premortem and candidate forecasts awaiting approval. This is what the log publishes.

Daily · Monday

7 September 2026

0 forecast updates

A company called PlexusDx published sixty-six product-comparison articles in two days, after publishing none the week before. It sells the products it compares. It offers custom-made copies of weight-loss drugs like Ozempic from about $129 a month, and custom-made versions of experimental drugs called peptides — short protein fragments sold for recovery, sleep and skin — from about $125 a month, with doctors licensed in all fifty states.

The US Food and Drug Administration has scheduled a public meeting for 17 September on giving testosterone to women going through menopause. Women are prescribed testosterone today, but no version is officially approved for them, so it is borrowed from men's products or made to order at a specialty pharmacy. The meeting notice says it will cover what evidence a company would need to get a women's version approved. That is the first visible outline of a proper route.

Quiet day on the numbers, loud day on the sourcing.

Read the brief

Read every brief139 published · since 19 April 2026

IIITrack record

The log is public. The Brier is the Brier.

Resolved — 4 backtests + 3 live

6/7

Direction correct (outcome matched higher-probability side)

0.146

Average Brier score · lower is better, < 0.10 is strong

21mo

Average lead time from cutoff to resolution

2/7

Contrarian calls · divergent from market/media consensus at cutoff

Audit begins

Q3 2026

Coverage — the fraction of resolvable events in our domains that Barque forecasts. Audit opens once the log has a full quarter of continuous re-evaluation behind it.

LiveForecasts resolved · 3

Resolved · 2026-06-10

compound-peptide-property-live

By 2026-07-31, The Compound has at least two peptide-adjacent properties live on production domains (comparison site, safety tool, skincare comparison, tracker app, or regulatory tracker)

Called

75%

Outcome

true

Brier

0.063

Resolved TRUE on the broader portfolio-ship interpretation per operator clarification 2026-05-15. Within the 103-day window (made 2026-04-19, horizon 2026-07-31): HRT Picks (hrtpicks.vercel.app, shipped 2026-05-12), glp1pets.com (shipped 2026-04-26), and Titrate all live, plus the public /numbers page on the mother site. Two-properties bar cleared with room to spare; functionally true ~51 days before horizon. Flagged for operator resolution 25 consecutive briefs (2026-05-16 through 2026-06-10) before this write. Validates the AI-speed-parallel-build thesis. Brier 0.0625 (forecast 0.75, outcome TRUE).

Resolved · 2026-07-26

BPC-157

PCAC does NOT recommend Category 1 (unrestricted compounding) for BPC-157 at the 2026-07-23 meeting; outcome is Category 2 return, deferral, or partial/conditional status short of full Cat 1

Called

70%

Outcome

false

Brier

0.490

Resolved FALSE. On 2026-07-23 the PCAC voted 8 to 6 with 1 abstention to RECOMMEND BPC-157 for addition to the 503A bulks list, i.e. the Cat-1-equivalent outcome the forecast said would not happen. KPV and TB-500 carried the same 8-6-1 tally and MOTS-c passed 7-5-2, so all four nominated peptides were recommended. Ground truth verified against bestpeptideforthat/src/data/peptides.ts pcacVote fields, which were primary-sourced at publication. THE LESSON, and it is the expensive one: the forecast reasoned that 'Regulator agent dominates regulatory decisions' and that FDA default on unapproved peptides without a US safety dossier is not-Cat-1. FDA staff DID oppose all seven substances. But the PCAC is an ADVISORY COMMITTEE, not the regulator, and it overrode its own staff. The model conflated the advisory body with the decision-maker. Future regulatory forecasts must model the advisory vote and the agency decision as TWO separate events with different actors and different base rates. The binding FDA decision is still pending and is a separate, unforecast question worth opening.

Resolved · 2026-07-26

pcac-july-multi-peptide

At least 2 of the 4 peptides reviewed at the 2026-07-23 PCAC meeting (BPC-157, KPV, MOTs-C, TB-500) receive a Category 1 or conditional Category 1 recommendation

Called

55%

Outcome

true

Brier

0.203

Resolved TRUE. All four peptides reviewed on 2026-07-23 (BPC-157, KPV, TB-500 at 8-6-1; MOTS-c at 7-5-2) were recommended for the 503A list, clearing the 'at least 2 of 4' bar with margin. This forecast was a deliberate upward revision made 2026-04-19 after the 12-peptide Category 2 removal wave, on the reasoning that the reclassification trajectory was more aggressive than first modelled. That revision was correct and it is the one that paid: the same session's BPC-157 forecast, which kept the conservative not-Cat-1 view, resolved FALSE at Brier 0.49. Lesson: when a trajectory signal contradicts a standing institutional prior, weight the trajectory. Resolved 3 days after horizon because the resolution loop had not run since 2026-06-10.

HistoricalBacktests · 4

Cutoff · 2022-08-31

Wegovy mainstreaming

GLP-1-class drugs become #1 US weight-loss Rx by 2024-12-31

Called

80%

Outcome

true

Brier

0.040

Novo raising guidance while supply-constrained was the cardinal signal. When an incumbent raises guidance under rationing, demand signal is maxed.

Cutoff · 2021-03-31

Contrarian

Clubhouse hype cycle

Clubhouse does NOT become top-3 audio platform by 2023-06-30

Called

75%

Outcome

true

Brier

0.063

Agent divergence capped signal strength. 3 of 5 agents negative override loud 2-agent positive (capital + celebrity). Incumbent clones shipping during peak hype is a kill signal.

Cutoff · 2022-12-31

Contrarian

NMN regulatory non-ban

NMN remains widely available despite FDA exclusion letter

Called

80%

Outcome

true

Brier

0.040

FDA warning letters without injunction or court action historically do not remove supplements from market within 24 months. Industry lobby + litigation + consumer demand compound.

Cutoff · 2023-03-31

BPC-157 Category 2 placement

FDA places BPC-157 in 503A Category 2 by 2023-12-31

Called

65%

Outcome

true

Brier

0.122

Weak signal environment. Protocol correctly expressed lower confidence via probability. Self-calibration working — weaker signal → wider error bars.

IVMethodology

Not a time machine. A calibrated one.

Barque isn't trying to predict everything. It's a domain-specific forecasting loop with a Brier score — the same primitive Tetlock's superforecasters use, scaled across internet signals. Five steps, every session, every forecast.

Observe

Pull signals from 20+ public sources across regulatory, financial, search, social, and cultural data feeds. Breadth first, never autonomous.

Cluster

Group items by emerging entity. A new drug name, a new retailer withdrawal, a new vocabulary on Reddit — cluster the signal, not the noise.

III

Simulate

Five agents reason independently: the Regulator, the Incumbent, the DTC Marketer, the Consumer, the Investor. If four rhyme, the signal is real. If they diverge, it's weaker than it looks.

Score

Signal Strength = volume × cross-domain × velocity × novelty × causal path. Below 20, we refuse to predict. Above 100, we predict with conviction.

Resolve

Every forecast has a date. When it passes, we score it with Brier — (probability − outcome)². The log is public. Corrections happen even when they hurt.

The ship we sendinto the signal,returns at dawn.

Predictions with resolution dates.

The ship returns. We log what it saw.

The log is public. The Brier is the Brier.

Wegovy mainstreaming

Clubhouse hype cycle

NMN regulatory non-ban

BPC-157 Category 2 placement

Not a time machine. A calibrated one.

Observe

Cluster

Simulate

Score

Resolve

Get the log before the market does.

The ship we send
into the signal,
returns at dawn.