Skip to content
7.1 — SOURCES

Where the numbers come from

The primary sources behind every figure on this page: official statistics, company filings and market data, each with its date and link. Every section also links its own sources.

How to cite · DOI → · Put the live score in your article → · SHA-256 · check the seal →

Full list of sources
  1. CNBC · market Nov 13, 2026
  2. FRED · S&P 500 · DJIA
  3. The Fed · decision of Sep 16, 2026
  4. Morningstar · Asset Owner Perspectives 2026
  5. Schroders · Global Investor Insights 2026
  6. Goldman Sachs · AI investment
  7. IEA · Powering AI
  8. World Bank · Commodity Outlook, Oct 2026
  9. Global X · Outlook 2027
  10. Similarweb · Investing, Oct 2026
  11. Bloomberg · ASKB
  12. BCG · Investor Survey 2026
  13. McKinsey · Investor Survey 2026
  14. AIMA · AI in the front office, Nov 12, 2026
  15. Gartner · AI spending 2027
  16. Crunchbase · venture Q3 2026
  17. InvestmentNews · ETF flows
  18. Kronos (AAAI 2026) · TSFM in finance
  19. SEC · EDGAR API · FRED API
  20. NYSE · data redistribution
  21. LiteLLM · routing
  22. QuantConnect · look-ahead bias
  23. Morningstar · asset owner survey, 2026
  24. Wells Fargo Investment Institute · commentary
  25. State Street · ETF flows to Oct 2026
  26. iShares · global ETP flows, Oct 2026
  27. World Gold Council · gold ETF flows
  28. IEA · energy investment · Energy and AI
  29. Federal Reserve · FEDS 2025-044 (LLM look-ahead)
  30. ALFRED · data vintages · SEC · search filings
  31. Bloomberg · ASKB agents · AlphaSense · Deep Research
  32. skfolio · QuantLib · OpenBB docs
  33. Charts: TradingView Lightweight Charts™ 5.3.0 (Apache-2.0) · replay series: FRED DGS10 · PCOPPUSDM · IR14270 · World Bank Pink Sheet
7.2 — ARCHITECTURE

From question to evidence

Five auditable layers. Every step leaves a record.

Question human · monitor US Claude · GPT · Gemini · Grok… China DeepSeek · Qwen · Kimi… Europe Mistral Le Chat Judge and fact-check facts ↔ primary sources weight: family × accuracy data: FRED · SEC · WB Answer + disagreement map web · app · Telegram Accuracy Ledger dated forecast → outcome → score API and archive “what the world’s AIs thought at moment T” accuracy shapes the next council
The five-layer research cycle
01 · SENSE

Radar

Track events, reports, macro data, flows and alternative data; link each fact to assets, sectors and supply chains.

02 · GROUND

Evidence

Extract figures from filings (SEC XBRL) and data releases. Preserve the source document, publication date and exact quote or figure. Flag stale, disputed or unverified claims.

03 · DEBATE

AI council

Models in the US, China and Europe work independently; an advocate, skeptic and arbiter test the case, and the case is run against history before the verdict (how).

04 · SIMULATE

Lab

Test historical regimes, factors and stress scenarios on point-in-time data, without future knowledge.

05 · ACT

Workbench

Every decision keeps an audit trail. Track scenarios, triggers, alerts, watchlists, notes and decisions. Revisit each call as facts change. Oversight pack: a decision and date T → the documents, versions, calculations, manual edits and sign-offs → a PDF/JSON with hashes and links for a client, a committee or a regulator. Free — a sample pack, rebuilt daily from the Ledger; Pro — your own work; Desk — the team.

OVERSIGHT PACK AS OF DATE T · SAMPLE, REBUILT DAILY

Loading the sample pack…

Sources: SEC EDGAR + FRED + licensed market data → fact graph + model council → research brief, scenarios, ledger and API.

7.3 — WHAT’S INSIDE

What runs inside Money

Each component, what it powers in Money, its license and its status. “Works” appears only when the engine’s acceptance tests for that module have passed; the date of the run is shown next to it.

ComponentWhat it powers in MoneyLicenseStatus
TradingView Lightweight Charts™ 5.3.0 ↗The replay chart: date T, the future hidden until you reveal it; next, the Lab chartsApache-2.0 · TradingView attribution
FRED ↗ (Federal Reserve Bank of St. Louis)Replay series: 10-year Treasury yield (Federal Reserve Board), copper since 1992 (IMF); the BLS gold import price index as a cross-checkFRED terms; each series keeps its owner
ALFRED ↗What was known on date T: macro series as first published, before revisionsFRED terms
World Bank Pink Sheet ↗Gold and copper prices in dollars since 1960 for the early episodesCC BY 4.0
edgartools ↗Report Checker: a figure → the SEC filing (accession number, acceptance time) → the exact passageMIT
skfolio ↗Portfolio Guard: CVaR and each holding’s contribution to riskBSD-3-Clause
statsforecast ↗The baseline every council forecast must beat: naive, seasonal naive, ETSApache-2.0
Polymarket and Kalshi APIsBrier score of the market itself; the divergence radarAPI terms; data kept local
Accuracy Ledger (built in-house)Sealed forecasts: hash, independent timestamp, the model’s training cutoff, Brier only on resolved questionsMoney’s own
QuantLib · QuantStats-format report (built in-house) · Chronos-2 · Qlib Alpha158 features + LightGBM (features computed in-house from Qlib’s definitions) · paper account (built in-house; Alpaca paper optional) · statsmodels · DuckDB + networkx · ccxt · broker note audit (built in-house)Bond math for the Report Checker and product payoffs for Portfolio Guard · backtest report · a challenger to the baseline · the Strategy Backtest’s machine-learning bar · paper account for the Broker Agent · skill or luck: factors, leverage, deflated Sharpe · Sector Researcher value chain: a chain and a date → reported spend, orders, revenue and cash flow, each dollar counted once · coin quotes across exchanges for the Divergence Radar’s “one asset, two prices” · Report Checker: upload a broker note → each figure and claim matched to a dated primary source → an annotated copy for analyst reviewBSD · Apache-2.0 · Apache-2.0 · MIT + MIT · MIT · BSD · MIT + BSD · MIT · Money’s own
AI-capex meter (built in-house on SEC companyfacts and 424B2/424B5 prospectuses)The Lab, AI capex: capex against operating cash flow over the same four quarters, new US-dollar notes from final prospectuses, and a status from fixed thresholdsMoney’s own · SEC public data
Narrative check (built in-house on edgartools)Report Checker: the company’s words in its earnings release (8-K, Ex. 99.1) against its 10-Q/10-K XBRL — confirmed, contradicts the report or not checkedMoney’s own · MIT
Council independence (built in-house)Council 10+: how many independent texts and positions the seats really give — TF-IDF distances, participation ratio, verdict entropyMoney’s own
Router prior (built in-house on Vals AI Finance Agent v2.1)Router: starting seat weights from the benchmark; the Accuracy Ledger’s own Brier takes over as questions resolveMoney’s own · benchmark figures cited
Patterns Money rebuilds from scratch — and what it does not run
ideas taken, code not used
  • TradingAgents ↗ — the bull / skeptic / risk-officer roles and debate rounds behind each council seat.
  • LiveTradeBench — a forecast sealed with a hash and a time before the outcome. PolyForm Noncommercial license, so the Accuracy Ledger is written from a clean sheet.
  • freqtrade — look-ahead and warm-up checks in every backtest. GPL, so only the idea.
  • vectorbt — the parameter heat map. Commons Clause, so only the idea.
  • FinSight — claim statuses “confirmed / contradicted / no support”. GPL, so only the idea.
  • FinRL-X ↗ — one target-weight contract from backtest to paper to broker, behind the Strategy Backtest’s execution gate. Apache-2.0, but a framework that sets the architecture, so only the pattern.
  • OpenBB ↗ — every number carries its provider, time and source.
not running in Money today

LEAN and NautilusTrader (heavy execution frameworks), Kronos (after Chronos-2), TimesFM 3 and Moirai 2 (noncommercial weights for Moirai), LangGraph, LiteLLM, vLLM and Ollama (when a client needs local models), FinGPT, FinRL, FinRobot, Hummingbot — research Money compares against, not components it runs.

LEAN ↗ · Qlib ↗ · Kronos ↗ · LiteLLM ↗ · vLLM ↗ · AI4Finance ↗

EVIDENCE AND METHOD: AI IN MONEY MANAGEMENT

What was measured about AI in money management, each line with its date, source and kind of evidence.

MONEY

See what moves markets

Markets worldwide. 10+ AI models, their independence measured on every poll. A view you can challenge, trace to sources and check against the public record.

The whole of Money on one page: every chapter, dataset, explanation and source. See everything →

Important disclosures

Money gives analytics and probabilities, not personal investment advice; figures carry their source and date, illustrations are marked as such, and past outcomes do not guarantee future ones.

All 14 disclosures
  1. Money provides analytics and probabilities, not individual investment advice; for personalized recommendations in the US and EU the Goal Navigator works through a registered investment adviser (RIA) or a licensed partner — as Betterment and Wealthfront are set up.
  2. The top strip shows Nov 13, 2026 closing quotes from open sources. Product data updates live.
  3. The council screen captures the real council answer of Nov 14, 2026; the Pulse ranking refreshes every hour.
  4. GitHub star counts and private-company revenue are third-party estimates; ranges are shown where available.
  5. Historical outcomes and scenarios do not guarantee future results. Interface illustrations are not trading signals.
  6. The “10-year above 5%” episodes and historical regimes use rounded reference data. The product calculates them from dated series.
  7. The Scenario Lab uses real facts to show the format; probabilities come from the live council.
  8. Model, data and broker connections require a terms and licensing review at launch; every model and data connection has a fallback.
  9. Survey figures describe their own samples: Morningstar covers 500+ institutional asset owners in 11 countries, not all investors. State Street flows cover the ETFs in its sample; the World Gold Council figure covers ten months (Jan–Oct), not a full year; the IEA’s $3.4tn is industry spending, not fund flows.
  10. The 2027 capital map is a working map, not a forecast. Forecasters disagree, and rising investment in a sector does not guarantee returns on its stocks.
  11. The Thesis Monitor card is an illustration assembled from facts on this page, not a recommendation on grid-equipment stocks.
  12. Web search is not access to all data: licensed research and expert calls are not on the open web. Open-source code does not grant rights to financial data.
  13. No public evidence shows a general-purpose LLM consistently out-forecasting markets.
  14. Word and deed records what public figures said and disclosed, with a short verbatim quote, its source and date, a link and a neutral result — not a judgment of anyone’s honesty or intent. Numbers about named people and funds come from primary filings and their own letters; interpretations are labeled “Money’s assessment”; Money is not affiliated with or endorsed by anyone it names, and none of them takes part in it. Commentators who are not public figures appear under a type label (“video commentator 7”) until they choose to be named; a disclosed position may be partial or out of date.