Serving status and badges mirror the platform's own vertical registry. A roadmap vertical shows the shape it would take and nothing more: it never renders a figure without an L0 badge.
| Vertical | What is behind it | Level |
|---|---|---|
| Private Credit | Disclosed-impairment score · out-of-filer LOFO AUC 0.59 [0.53, 0.65] strict, 97 events / 0.69 [0.67, 0.72] union, 545 events (not a probability of default) · as of 2026-11-09 | In-Sample Real · Unvalidated |
| Macro & ESG | Fed and ECB statement tone · scored clause by clause from each bank's own statement, with its date and source · the tone → next-decision test does not beat the last-move rule (honest negative, no accuracy claim) · as of 2026-10-22 | Real · Unvalidated |
| Manufacturing | MF7 nowcast · +3.6% vs the prevailing mean, not significant (DM p 0.45) · no accuracy claim · as of 2026-10-22 | Real · Unvalidated |
| Commodities | COT positioning · CL OOS AUC 0.6137 (n=140) — not significant against a coin across 22 markets (honest negative) · as of 2026-10-22 | Real · Unvalidated |
| Fixed Income | Treasury 10Y one-day VaR · next-day loss beat it on 1.09% of 6,699 days (target 1%, 2000–2026) · calibrated, but no better than a rolling normal VaR (no accuracy claim) · as of 2026-10-22 | Real · Unvalidated |
| Equities | SEC XBRL fundamentals for 12,826 companies (filed to 2026-08-03) · 13F institutional holdings for 33,449 securities (quarter 2026-03-31) | Real · Unvalidated |
| FX | REER misalignment · valuation context only — the reversion claim was tested on real data and does not beat a random walk (honest negative) · as of 2026-10-20 | Real · Unvalidated |
| Energy | NY Harbor 3-2-1 crack · observed margin — the reversion claim was tested on real data and does not beat a random walk (honest negative) · as of 2026-11-03 | Real · Unvalidated |
| Private Equity | Real data: private-fund fundraising from SEC Form D, the SBIC roster, sponsors' 13F holdings, marks and non-accruals on sponsor-backed loans across the BDC panel, and a target company's filed fundamentals with its debt capacity at today's cost of debt — each figure with its source and data date. Fund IRR/TVPI need paid cash-flow data and show a dash with the reason | Real · Unvalidated |
| Crypto | Next Bitcoin difficulty adjustment · 80% band held 79.7% of 330 epochs; error 53% below "no change" (L3) · as of 2026-10-20 | OOS Validated Real · Unvalidated |
| Logistics | BTS freight index (TSI) with its data date · one-month band is an honest negative (80.7% coverage, misses cluster) · as of 2026-10-20 | Real · Unvalidated |
| Infrastructure | US highway traffic, trailing 12 months (FHWA, observed) · the next-month nowcast is published as an honest negative: worse than a rolling AR(1) on year-on-year growth (L1) · as of 2026-10-22 | Real · Unvalidated |
| Insurance & ILS | Median carrier loss ratio · observed — the state 1-in-10 level of NWS-recorded storm damage was tested on real data: an honest negative, served as a historical statistic · as of 2026-11-01 | Real · Unvalidated |
| Banking | Real data: every US insured bank's capital, asset quality, earnings, liquidity and CRE concentration from FDIC and FFIEC call reports, with peer percentiles; offices by market; a failure-risk screen tested on every FDIC failure since 2008 (no accuracy claim: one ratio alone ranks failures better); rate sensitivity against FRED. Uninsured deposits, charge-offs and deposit shares by market are not loaded and show a dash with the reason | Real · Unvalidated |
14 business development companies, read from their own SEC EDGAR Schedules of Investments. Public filings, public tickers.
The operational database is built from public primary sources — government statistical programs, regulatory filings and market data. Every figure on this site carries the date it was measured.
Schedules of Investments are parsed from EDGAR directly. Where a filing omits a field, the field is recorded as undisclosed rather than filled in.
Macroeconomic inputs, such as the Treasury yield curve and unemployment series the recession models are tested on, come from published official series. No recession probability is shown today: no recession model has passed its test.
FRED (latest revisions), ALFRED point-in-time vintages for selected series, Treasury curves, BIS, ECB, global rates panels
SEC EDGAR (13F, Form D, BDC XBRL, fails-to-deliver), FDIC, FFIEC, NCUA, CFPB
CFTC Commitments of Traders (the futures markets in the commodities registry), commodity price programs, corporate bond credit conditions (Federal Reserve Board GZ spread and excess bond premium; New York Fed corporate bond market distress index)
BLS CES/CPS/JOLTS, Census ACS, county-level employment (QCEW), migration flows
FHFA HPI, Zillow research series, building permits
NOAA/nClimDiv, drought monitor, HURDAT2, BTS freight, supply-chain pressure indices, web & social signals
Ask for a past date and you get only what had been published before that day, or a refusal. Fixed income, equities, FX, logistics, commodities and private equity answer from data published before the date you ask for: company filings and Form D notices by their filing date, COT reports by their Friday release, market prices from the next day (EIA spot prices after EIA's weekly release), and monthly and quarterly statistics by their publisher's release calendar, at their current revision (the first-release value is not stored). Every other vertical refuses a past date, naming the data it cannot date, rather than show today's numbers.
You have the roster and the levels. The validation record behind them is one page away, along with thefour claims we withdrew.