Skip to content

For those who follow theses and markets

3.1 — WHAT MATTERS NOW

November 2026. What really matters?

AI is hitting the balance sheet. Investors must weigh its potential against borrowing costs and real-world bottlenecks: power, copper, chips and capital.

CAPEX
$700bn+

2026 capex by the largest data center operators (IEA); Goldman puts global AI investment above $1tn.

IEA ↗ · Goldman Sachs ↗
THE BIG TEST
73%

of asset owners question the payoff from AI valuations and capex; 76% cite inflation (Morningstar 2026).

Morningstar ↗
INVESTOR GOAL
84%

prioritize diversification; 85% expect more volatility (Schroders, respondents managing $72tn).

Schroders ↗
METALS
+25% → −3%

World Bank sees metals prices at +25% in 2026, then −3% in 2027. One forecast won’t capture that swing.

World Bank ↗
The edge. Better decisions start with better tests. Money runs each thesis through 10+ models whose independence the engine measures, primary data, a skeptical debate and historical analogs. You see what would make the thesis hold or break.
More context: the core question, what asset owners flag, the investor pain map
what Money checks behind every position

You enter a company, sector or asset → Money checks how much future profit the price already assumes → you get a dated verdict: still underpriced, or already paid for.

Money does it one company, sector or asset at a time: the evidence, the counter-case, the scenarios and a dated record of what changed. Wells Fargo’s November outlook circles the same issues — data-center payback, how the build-out is financed, rising leverage and whether demand turns into cash flow.

Wells Fargo Investment Institute ↗

what asset owners call material · Morningstar 2026
  • Inflation and whether AI valuations and capex pay off — the card above
  • 71% — the rise of generative AI
  • 66% — energy and logistics crises

500+ asset owners in 11 countries · Morningstar ↗

Investor pain pointWhat the investor needsWhere Money answers
Fear of a large lossKnow how much could be lost — and whyPortfolio Guard: stress tests and historical regimes
Information overloadA few material events, not thousands of headlinesThesis Monitor: alerts only when the facts change
Distrust of data and adviceThe primary source, the calculation and its limitsEvidence chain and Report Checker
Fear of missing outTell a real opportunity from hypeCouncil debate: advocate, skeptic, arbiter
Hidden concentrationSee when different holdings ride on one factorPortfolio Guard: factor exposure
Money beyond Wall StreetKnow what a local asset will really be worth in the currency I will spend — after the exchange rate, local tax and inflationGoal Navigator: My money in the world
Wanting a machine to do itHand the routine to AI without losing the brakes — and see its record before trusting it with moneyBroker Agent within your limits; the Accuracy Ledger first
No timeStay on top of positions without reading the news all day30-minute routine and morning brief
UncertaintyA plan for more than one outcomeScenario Lab: base, upside and downside cases

Behind the map: investors target 7.9% a year, yet 22% of their portfolios sit in cash (Fidelity — the yardsticks behind Goal Navigator), and only 44% feel confident about a comfortable retirement (Betterment) — they want growth but hold back for fear of a loss. Individual investors and small funds often see news after algorithmic traders; a single model can invent a valuation multiple or repeat a stale quote; and when a handful of AI companies dominate an index, a day like Nov 5 hits portfolios that looked diversified. McKinsey cites geopolitics, AI’s impact and capital discipline; Janus Henderson cites worries about AI valuations, distrust of recommendations and how hard they are to verify. McKinsey ↗ · Betterment 2026 ↗ · Fidelity 2026 ↗

This week’s Pulse: what moved markets, the events ahead, themes on the radar
pulse · edition of Nov 14, 2026

The AI boom meets a 5% Treasury yield

  • Earnings vs. rates — S&P 500 Q3 earnings growth near 30% (FactSet ↗), while the 10-year crossed 5% for the first time since 2023 and hit 5.16–5.18% in September, the highest since 2007; on Nov 13 it closed at 5.15% (US Treasury ↗).
  • Nov 5 — after Democrats won both chambers, fears of limits on data-center construction sent the S&P 500 down 4.4% — its worst day since April 2025 — and the Nasdaq-100 down 6.4%, its worst since March 2020 (HNGN, Nov 5 ↗).
  • Inflation, Nov 10 — October CPI rose 0.4% on the month (BLS ↗). The consensus was 0.3%, and Fed Governor Waller raised the possibility of a 50 bp hike (MarketWatch via Morningstar ↗).
  • Company filing, Nov 12 — Nvidia’s 8-K: on Nov 11 the board accepted Jensen Huang’s resignation as CEO and named CFO Colette Kress interim CEO (SEC filing ↗). The Thesis Monitor puts a filing like this on every thesis that holds the company; the next filing replaces this line.
  • Concentration — AI companies account for more than 80% of S&P 500 growth in 2026 and about half its market cap (J.P. Morgan AM). Morningstar put US stocks about 10% below fair value at Sep 30 — yet without seven stocks the market traded right at fair value (Morningstar ↗).
  • October rotation — Morningstar’s US sector indexes: energy +14.05%, consumer defensive +11.55%, utilities +9.87%, healthcare +8.82%; communication services −14.72%, where Alphabet and Meta make up 78% of the index (Morningstar, Nov outlook ↗). Index returns, not measured money flows; Money’s own rotation, country by sector, is the board’s Rotation view.
  • Mood, week to Nov 4 — AAII bearish sentiment 48.4% (up 6.9 points on the week) against a 31.5% average (AAII ↗).
  • Nov 13 close — S&P 500 7,561.67 (−0.40% on the day, +2.51% on the week, ending a four-week losing streak; record 7,844.67 on Oct 6). Dow 52,347.55 (+0.10%). AI, chip and momentum stocks sold off; defensives and the equal-weighted index held up (CNBC ↗).

CNBC ↗ · Market Scholars ↗ · Goldman Sachs AM ↗ · TMGM ↗ · Kitco ↗ · Wells Fargo ↗

event calendar · calls sealed before the event
Freethe whole calendar; each call’s hash and direction the moment it is sealed, everything after resolutionFreeevery thesis verdict and sealed point once it resolves$2the council’s ranges, probabilities and invalidation conditions before resolution$2central-bank wording shifts, release surprises in σ and prospectus amendments while fresh, then free (planned)Onealerts on your thesis: the service worker on your device matches each new event (a condition, a ticker, a filing) to the thesis kept in your browser and notifies you once per event; the server never sees your thesis
DateEventWhat the council seals
Nov 17Nvidia earnings, days after its Nov 12 leadership filing (SEC 8-K ↗)Data-center revenue vs. guidance; reaction of the chip index
Nov 18–19Anthropic IPO (Nasdaq: ANT) — 825mn shares at $70–78 — about $0.92–1.02tn on the ~13.09bn shares outstanding after the offering, up to ~$1.4tn fully diluted; pricing expected Nov 18, first trade Nov 19 (WSJ via Decrypt ↗ · SEC S-1/A ↗). The range is public; the council seals its own callPricing vs. range; first-week move
Dec 9Fed decision — after the Nov 10 CPI, markets expect a hike (Axios, Nov 10 ↗; live odds: Kalshi ↗, Polymarket ↗, CME FedWatch ↗); when the council seals its forecast, a dated snapshot of these odds is kept beside it and both are scored after the decisionHold, +25 bp or +50 bp, with probabilities
FED DECISION · MARKET ODDS BESIDE THE COUNCIL

Loading the Kalshi and Polymarket odds…

Dec 10November CPIAbove, at or below consensus
Dec 17ECB decision (ECB calendar ↗)Deposit rate held or changed, with probabilities
Dec 18Bank of Japan decision after its Dec 17–18 meeting (BoJ ↗)Hold or hike, with probabilities
In: a scheduled event and the question that matters for your thesis · Does: all 10+ models lock a probability and range before the release, with the condition that would prove them wrong · Out: after the event, the call is scored automatically in the Accuracy Ledger next to the simple benchmarks — and your Thesis Monitor shows whether the result touched a break condition.
More: what the calendar reads — other central banks, wording, σ surprises, prospectuses, rule changes

Other central banks: the decisions and CPI of every market on the board — the Bank of England, RBI and Copom after the ECB and the Bank of Japan — and new listings on HKEX and NSE are sealed the same way and scored market by market; RBI meeting dates come from RBI’s own schedule, CPI dates from the calendars of the BLS, Eurostat, IBGE, the ONS, the Statistics Bureau of Japan and MOSPI.

Central-bank wording: after each decision the engine compares the statement with the previous one — phrases added, dropped or softened — and sets that shift beside the sealed call and in the calendar row: the Fed, the ECB and the Bank of Japan first, then the central banks of the board’s markets. Free — the shift once the next meeting has passed; $2 — while it is fresh. Sites that render by script or answer bots with 403 are read now: the BIS feed of central bankers' speeches (50 speeches from 34 institutions, Oct 29 – Nov 12, 2026), the latest speeches of all nine regional Reserve Banks that were missing from the speaker map, RBI's policy corridor from its own daily release (repo 5.50%, SDF 5.25%, MSF 5.75%, Nov 16, 2026) and the ECB's statement of Oct 29, 2026 set against Sep 10: rates unchanged after a 25 bp hike, 54 sentences new, 61 dropped, 21 kept.

Speaker map: Live for the Fed Board, the ECB and the Bank of England, with dissents from the FOMC statements, and for the Bank of Japan’s board and the regional Fed presidents whose banks publish a speech feed (Dallas, Richmond, San Francisco); the other regional presidents are next. Every speech, interview and dissenting vote of each policymaker — Fed governors and regional presidents, the ECB Governing Council, the Bank of England’s MPC, the Bank of Japan’s board — read against that person’s own earlier words, so a hawk turning cautious stands apart from a hawk repeating himself; the map sits beside the sealed call for the next meeting, and the Accuracy Ledger later shows whose words led the decision. Free — the map once the meeting has passed; $2 — while it is fresh.

CALENDAR · SPEAKER MAP · ENGINE SLICE

Speaker map: each policymaker against his or her own words

CALENDAR · SPEAKER MAP · REGIONAL FED AND BANK OF JAPAN

Regional Fed presidents and the Bank of Japan board

Surprise in σ: the released figure, the median of the council’s sealed calls and the history of misses on that series → “came in at X, expected Y, N σ above” in the calendar row; above |z| of 2, with a condition touched, it becomes a Thesis Monitor event.

Prospectus amendments: each new S-1/A is compared with the previous version (edgartools) — price range, share count, debt, risk factors, dates — and the listing row shows what changed before the first trade, linking both filings. First case: Anthropic's second amendment (10 Nov 2026) set the first price range, $70–$78, and moved the risk that its models have behaved in unintended ways from 13th place to first (S-1/A ↗). Free — the day’s main release and the fact that a prospectus changed; $2 — every release and the full amendment read the moment they land, free a day after the release and after the listing; One — on your thesis.

Rule changes: the dates when rules for markets and for AI in finance take effect — exchanges, securities regulators, EU AI Act stages — from the official text, with withdrawals logged as events too: on June 12, 2025, the SEC withdrew 14 proposals, including the predictive data analytics rule for brokers and advisers (SEC ↗). Each change is read against the previous text — what was added, dropped or reworded — with the publication date and the effective date kept apart; a proposal never changes the live Market Profile, and a repeal is logged as a new version. Free — the decision, its dates and the official link; $2 — the full change map while fresh, open to all after 30 days; One — alerts when a rule touches your thesis.

Government bond auctions: Live for the US, Germany and Japan from each debt office’s own results. Each sovereign auction in the 20 markets of Government bonds of the world → bid-to-cover, the tail against the pre-auction yield, the foreign share and the surprise in σ against that issuer’s own history. A bond-vigilante flag lights when a weak auction, a rising term premium and a falling currency land in the same week; the Atlas’s fiscal-limit flag stands beside it. Debt agencies’ borrowing plans and central-bank statements are read in the original language; a shift in a borrowing plan is $2 while fresh, then free. Free — the auction result; $2 — the σ surprise and both flags while fresh, open to all after 30 days.

CALENDAR · PROSPECTUSES · CENTRAL BANKS · σ · RULES

What the calendar reads

Anthropic, PBC · S-1/A No. 2 · 10 Nov 2026
price rangeshares offerednet proceedstotal debtvalue at midpoint
$70.00–$78.00825.0 mn$84.6 bn$6.2 bn$968.4 bn computed

2026-11-18 · Pricing of the Global Offering · 2026-11-19 · Listing on the Nasdaq Global Select Market · FWP ↗

S-1/A No. 1 (2 Nov) → S-1/A No. 2 (10 Nov) · previous filing ↗ · new filing ↗

  • first price range: $70.00–$78.00; 801.4 mn new shares and 23.6 mn from selling stockholders; option for 123.75 mn more
  • concurrent private placement of Class N stock: $26.0 bn
  • risk factor reworded and moved from #13 to #1: «Our models have in the past behaved, and may in the future behave, in unintended, misaligned, or autonomous ways…»
CALENDAR · NEW LISTINGS · NSE

New listings on NSE

CALENDAR · CPI DATES · UK, JAPAN, INDIA

CPI dates: UK, Japan, India

Fed · FOMC calendar ↗ · BLS · CPI schedule ↗ · CNBC ↗

Theme on the radarWhere it stands, Nov 2026Where Money tracks it
AI capex vs. borrowing costsThe organizing story: capex lifted chip stocks “impressively” while fears of AI “displacement” weighed on software (Columbia Threadneedle, 2027 outlook ↗)Scenario Lab · AI-capex meter
Elections and the FedData-center regulation risk after the Nov 3 midterms; after the Nov 10 CPI, markets expect a hike on Dec 9 (Axios, Nov 10 ↗)Event calendar
Robotics and Chinese assetsChinese industry and supply chains as the next leg of AI demand: China installed 295,000 industrial robots in 2024, 54% of the world total (IFR World Robotics, Sep 2025 ↗)Country × Sector board
Copper — “AI’s new oil”Records of $14,533/t on the LME and $6.95/lb on COMEX in September on AI and grid demand (Yellow ↗)Scenario Lab · copper
Gold under rate pressureFrom a record $5,595.47 intraday spot on Jan 29 (World Gold Council; helped by central-bank buying in China, India and Poland) to ~$3,881 on Nov 13; −2.46% on Nov 13 alone. 2025 was gold’s best year since 1979 (WGC ↗)Scenario Lab · gold
Event odds as a marketKalshi and Polymarket handled $50.6bn in July 2026; Kalshi set a $15.27bn weekly record in September (Tangem ↗)Divergence Radar
Oil and geopoliticsBrent at $95.84 at the Nov 13 close amid the Iran war (Vista Partners ↗); geopolitical conflict tops the underestimated risks in McKinsey’s survey of 112 investors (McKinsey ↗)Portfolio Guard · “oil at $120”
Defense, space and drones$35.6bn of venture funding into defense and defense-adjacent startups in H1 2026 (PitchBook via Tectonic ↗); SpaceX’s $75bn IPO on Jun 11 (NPR ↗)Crowding Index · themes
Bitcoin~$67–69k on Nov 13, −30% year over year, with outflows from spot ETFs (CNBC ↗)Pulse
Beyond AIGLP-1 and longevity drugs: Eli Lilly’s Q3 2026 tirzepatide sales reached about $16.1bn (Mounjaro $10.7bn, Zepbound $5.4bn) (Pharmaceutical Technology ↗)Report Checker
2027 watchlistsGlobal X: AI, defense and space, infrastructure and automation, energy, critical resources. Grid equipment (Eaton, Schneider Electric, Siemens Energy), copper (Freeport-McMoRan), uranium (Cameco), SMRs, transformers and cables, quantum-resistant cybersecurity, satellite networks, Dow dividend payers with real free cash flow (Global X, 2027 outlook ↗)Accuracy Ledger · house forecasts
Concentration and AsiaThe 10 largest stocks are more than ¼ of MSCI World; Asian emerging markets have led this year. Earnings growth is broadening: for Q3 2026 the Magnificent 7 show earnings growth of 22.6% vs. 32.6% for the other 493 S&P 500 companies — blended rates (reported results plus estimates for companies yet to report) as of Nov 3, and growth rates, not shares of total profit (FactSet ↗)Portfolio Guard · hidden concentration

Columbia Threadneedle 2027 ↗ · Semafor ↗ · Rio Times ↗ · Yellow ↗ · Tangem ↗ · DeFi Rate ↗ · McKinsey ↗ · Vista Partners ↗ · Global X ↗ · World Gold Council ↗ · AP · gold’s best year since 1979 ↗ · FactSet ↗

Research as of Nov 14, 2026: where capital went in 2026 — on Wall Street and beyond it — and where it may go in 2027 · government bonds of the world
capital flow atlas · dated study · as of Nov 14, 2026
Freethe study, its conclusions and sources$2full tables, period comparison, series historyContractdata export

Pools of money, kept apart on purpose: they have different denominators and cannot be added into one “global investment” total.

PoolPeriod and measureSource
US ETF flowsRecord $1.54tn by September, $1.75tn by October. Bond ETFs a record $469bn; technology led sectors at $66bn; dividend and income equity ETFs more than $100bn. US equity funds ~$974bn in the 12 months to October vs. ~$201bn for developed markets ex-US. Demand for rates, corporate credit and EM debtInvestmentNews ↗ · ETF Trends ↗ · State Street ↗ · iShares ↗
Fund flows of Vanguard, Invesco, Fidelity and VanEck ETFs · SEC N-PORTQuarterly, about 60 days late. Flow (sales minus redemptions) and net assets are separate figures. Vanguard rows cover the whole fund, all share classes.SEC EDGAR N-PORT ↗
Global ETFsA record $2.37tn of net inflows in 2025 and $19.85tn of assets at year-end (ETFGI); $20.64tn by the end of January 2026. In Q1 2026 active ETFs took $179.3bn against $178.7bn for passive; equity ETFs $162.5bn, bond ETFs $112.2bn of a $358.1bn total (S&P Global). Providers differ in coverage and method: ETFGI puts Q1 net inflows at a record $626.42bn, so the two estimates are never added or read as one series — and assets, a stock, are never added to flowsETFGI, S&P Global Market Intelligence
Managed assetsThird-party managed assets a record $151tn at the end of 2025 across 42 countries; net flows equal to 3.2% of opening assets, concentrated in a small group of managers — a stock of assets, not a flowMcKinsey, Oct 2026
Gold ETFsAbout $41bn into physically backed funds worldwide, Jan–OctWorld Gold Council
VentureAI took $102bn, 64% of global venture funding, in Q3 (Crunchbase) and 82.7% of US deal value Jan–Sep (PitchBook-NVCA); global startup funding $685bn in 9 months vs. $449bn in all of 2025 (Dealroom)Crunchbase, PitchBook-NVCA, Dealroom
Private equity$1.5tn across 14,650 deals in the first nine months of 2026, $854.6bn of it in the Americas (6,845 deals), $521.4bn in Europe, the Middle East and Africa (6,249) and $137.8bn in Asia-Pacific (1,109) — fewer, larger deals. By sector: technology, media and telecom $479.1bn, energy and natural resources $243.3bn, industrials $122.6bn; corporate carve-outs $182.2bn. Exits were worth $985.2bn; the trailing-12-month exit count hit a 22-quarter low while exit value reached an 18-quarter highKPMG Pulse of PE, Q3 2026 ↗
Corporate and industry spendingGlobal energy investment ~$3.4tn (industry spending, not fund flows); five tech companies spent more than $400bn on capex in 2025, mostly data centers; Goldman expects AI capex to rise as a share of GDPIEA, Goldman Sachs
Asset-owner allocationsRaised 2026 allocations to private equity (43% of respondents), infrastructure (35%) and private credit (32%) — shares of asset owners raising allocations, not dollar flows. Wealthy individuals (600+ in the UK, Europe, Asia, the Middle East and Africa; Barclays Private Bank survey, Oct 1, 2026; percentages as reported by WealthBriefing): 60% plan to raise their private-market allocations, even though the average share they allocate to private markets edged down from 30% to 27%; within private markets, technology (63%) and healthcare (54%) are the sectors most want more of — shares of respondentsMorningstar, 2026 ↗ · Barclays Private Bank ↗ · WealthBriefing ↗
Private markets and private creditThree measures, never added together. Loans outstanding: private credit loans to US borrowers were $1.54tn at end-June 2026 (Fed Z.1, cost basis), up $185bn in a year. Assets under management worldwide, including committed capital: $1.7–2.1tn for 2023 (Fed, IMF). Private funds of SEC-registered advisers: $12.2tn net assets in private equity, venture, real estate and other private funds at end-2025 (Form PF). Free — the range row; $2 — holders, fund types and eight quarters.Fed Z.1 ↗ · SEC Form PF statistics ↗ · IMF GFSR ↗
BDC loan booksWhen the loans of the largest business development companies come due and which have stopped paying, read line by line from each BDC’s own Schedule of Investments. Today: 7 BDCs with $160bn of investments at fair value (Sep 30, 2026) — loans on non-accrual are 2.3% at cost and 1.3% at fair value; $47.9bn of their debt is due within three years; 13% of the debt pays part of its interest in kind. Fair value and face value are shown apart. Free — this row; $2 — every BDC with its maturity ladder and its largest non-accrual loans.SEC EDGAR 10-Q ↗
Private company share prices · Forge Price and the last funding roundOnly what Forge publishes openly, with its date. Forge Price is an indicative daily price; a funding round is a deal. Where Forge shows no figure, it is marked not measured.Forge ↗
Off Wall Street · money that moves outside the marketFirst sources only; each figure shows its measure (flow, stock or share), period, link and quote. A figure the source does not print is marked not measured.IFSB ↗ · PBoC ↗ · World Gold Council ↗ · World Gold Council ↗ · World Bank ↗ · IFAD ↗ · SEBI ↗
Who financed US real estateWho financed US real estate in a year: $729bn of net new mortgages — agency pools $314bn, banks $211bn, private ABS $136bn (Fed Z.1)Fed Z.1 ↗
Energy FDIEnergy FDI 2024: Australia +$13.6bn into mining and oil and gas, Canada +$11.9bn (OECD)OECD FDI ↗
Fund and manager holdingsHow the disclosed positions of large managers and funds changed, from their own SEC filings: managers’ 13F (quarterly, up to 45 days after quarter-end) and funds’ N-PORT (monthly holdings). Money counts the change in stakes net of price moves; every line carries the report date and a delay tag — these are positions as of a past date, not today’s flow: the SEC describes Form 13F as a report of holdings, not a trade history, so a change between two quarters is shown as a change in holdings, never as a purchase made on the filing day. The same fund holdings feed the look-through in Portfolio Guard. Today: seven managers’ latest 13F (Berkshire Hathaway as of Sep 30, 2026, public after 47 days) and five funds’ public N-PORT. ARK Innovation ETF opened SpaceX at 4.9% of the fund and left Roku (N-PORT of Jul 31, 2026); Berkshire roughly doubled its Lennar stake while Lennar’s price fell 9.8% over the quarter — such lines carry the flag.
13F · MANAGERS · NET OF PRICE

Managers' holdings (13F): change in stakes net of price

Loading the engine slice…

N-PORT · FUNDS · NET OF PRICE

Fund holdings (N-PORT): change in stakes net of price

Loading the engine slice…

More: faster disclosures, three dates, who filed

Faster disclosures join the same table: company insiders (SEC Form 4, within two business days of the trade), stakes above 5% (Schedule 13D, within five business days — the deadline since Feb 5, 2024, SEC release 33-11253 ↗), members of Congress (periodic transaction reports: House Clerk ↗, Senate eFD ↗), the daily holdings files some active ETFs publish (ARK: every trading day; the change from one day to the next is shown, not a reported trade) and, beyond the US, company executives’ own dealings (PDMR) under the EU and UK Market Abuse Regulation (within three business days). Every line carries three dates — when it was done, when it became public and when Money recorded it — and names the filer, not a person: a Berkshire Hathaway 13F (EDGAR ↗) shows Berkshire’s holdings, not Warren Buffett’s own trade, and a fund’s action is tied to a manager only when a filing shows that person’s own decision. A filing that shows more shares of a stock while its price fell is flagged; “bought” is used only for a filing that reports a trade (Form 4, Schedule 13D). Today the engine reads the last seven days: Schedule 13D and 13G from their structured EDGAR filings, Germany’s directors’ dealings from the BaFin database and UK PDMR notices from the FCA National Storage Mechanism — each line with the deadline it was due against (13D: 5 business days, amendments 2; MAR Art. 19: 3 business days).

13D · 13G · ACTIVE ETFs · MAR ART. 19

Faster disclosures: three dates on every line

Loading the engine slice…

These deeds feed the “said vs did” line of Word and deed and the two layers of a long-horizon investor’s card. Free — single positions with their date; $2 — full series of changes, open to all after 30 days; One — on your portfolio; Pro — your own selection
SEC 13F ↗ · SEC N-PORT ↗ · SEC 13F FAQ ↗ · SEC Form 4 ↗ · SEC 13D rule ↗ · MAR Art. 19 ↗ · BaFin directors’ dealings ↗ · FCA NSM ↗ · ARK daily holdings ↗

Beyond Wall Street · as of Nov 15, 2026. For most families the “market” is a flat and gold: financial assets are 25.8% of household wealth in India, about 52% in China and about 79% in the US (UBS Global Wealth Report 2026). Each layer below is a different measure — exchange value, contributions, holdings, turnover — so the layers are never added into one total.

LayerWhere it standsSource
Other exchangesMarket cap at end-September 2026: Asia-Pacific $57.5tn and Europe, Middle East and Africa $25.4tn, against $86.2tn for the Americas. Nasdaq $45.1tn, NYSE $33.6tn; Shanghai $10.2tn, Japan Exchange Group $8.9tn, Euronext $8.0tn, Shenzhen $7.1tn, Hong Kong $5.8tn, Taiwan $4.9tn, BSE and NSE $4.9tn each (largely the same companies — never added), Saudi Exchange $2.4tn. All listed companies worldwide: $151.9tn at end-2025WFE, Nov 2026 ↗ · WFE, 2025 ↗
Mass saving routesIndia: SIP contributions of ₹32,160 crore in October 2026, from about 10 crore contributing accounts in September (accounts, not people); NSE passed 13 crore unique investors in April. Japan: 28.26mn NISA accounts at end-2025, target 34mn by end-2027. Brazil: Tesouro Direto — 36.4mn registered, 3.83mn active, R$272bn held (August 2026); 60.6mn Brazilians invest (2025). Developing economies: 40% of adults saved in a financial account in 2024, 10% through mobile moneyAMFI via Kotak Neo ↗ · 5paisa ↗ · NSE ↗ · FSA ↗ · B3 ↗ · ANBIMA ↗ · World Bank Findex ↗
Property and goldAll real estate: $393.3tn at end-2024 — residential $286.9tn, commercial $58.5tn, farmland $47.9tn (a stock of wealth). Gold: 2,175 t of total investment demand in 2025, 1,374 t of it bars and coins; about 268 t of bars and coins in Q3 2026Savills ↗ · WGC 2025 ↗ · WGC Q3 2026 ↗
Pensions$68.3tn in the 22 largest pension markets at end-2025, 65.8% of it in the USThinking Ahead Institute ↗
State moneySovereign wealth funds above $15tn; all state-owned investors — sovereign funds, public pensions and central banks — about $60tn at end-2025. The largest single fund, Norway’s, held NOK 22,014bn at Sep 30, 2026 (NBIM ↗); ADIA about $1.13tn at end-2025 (Global SWF estimate). Allocation shifts: Live for NBIM (Norway), GIC (Singapore), CIC (China), CalPERS (California) and CPP Investments (Canada) from their own reports; PIF is next. NBIM, PIF, GIC and CIC, from their own reports → how each splits across listed equities, bonds, real estate, unlisted infrastructure and regions, and how that split moved since the previous report, with a delay tag. The same dated view can read public pension and multi-asset manager reports: target allocation versus actual holdings, commitments versus cash called, and liabilities due before an asset can be sold. CalPERS, March 31, 2026: public equity 5.0 pp under its interim target, private equity 4.5 pp over; its private-equity funds have called $130.2bn of $163.0bn committed (79.9%), $45.1bn not yet called; funded status 81.6%. CPP Investments, March 31, 2026: public equities 35.6% of net investments (28.3% a year earlier), unfunded commitments C$80.4bn (C$63.4bn). CIC, end-2024: alternative assets 48.5%, public equity 34.7% (+1.5 pp). Each change points to the fund’s own report; holdings, commitments and flows stay separate, with no ranking by assets.
PENSION AND SOVEREIGN FUNDS · OWN REPORTS

Pension and sovereign funds: target versus actual

Loading the engine slice…

Free — the conclusions and NBIM’s latest split; $2 — full series and country cuts, open to all after 30 days
Global SWF ↗
Off-exchange tradingForeign exchange $9.6tn a day and OTC interest-rate derivatives $7.9tn a day in April 2025 — turnover, not assetsBIS ↗
Government bonds of the worldFor your own date and currency: What stays with me in My money in the world. Here — the country map of the 10 largest government bond markets, then 20 (coverage, not a ranking), with a fiscal-limit monitor, who buys the debt, the maturity wall and index events. Free — the map and the top 10, the method and today’s interest burden, the latest holder facts, the largest maturity wall and the latest index event; $2 — the full table with history, all 20 fiscal limits, holder history, the maturity wall and index events, open to all after 30 days; One — a fiscal-limit alert at your own threshold
Country map, fiscal limit, who buys, maturity wall, index events

Coverage: the US, China, Japan, Germany, France, Italy, the UK, India, Brazil and Canada, then South Korea, Mexico, Spain, Australia, Switzerland, Poland, Indonesia, Saudi Arabia, South Africa and Turkey.

Country map: the 10-year yield and the 10Y−2Y slope, real yield, spread to US Treasuries and to the Bund, debt to GDP, interest cost as a share of budget revenue, foreign holders’ share, and weight in bond indexes by debt and by GDP — on one data schema, because local series are published on different methods (a zero-coupon point for Germany and the UK, a real yield for Brazil).

Fiscal-limit monitor: debt, deficit, average maturity and the maturity wall → the 10-year yield at which interest would pass a stated share of budget revenue, the buffer to it in basis points and a flag: normal, watch, limit. It is a scenario at that threshold: the threshold value and the source of the budget data stand beside the flag; One sets its own threshold.

Who buys the debt: TIC, central-bank balance sheets, non-residents, stablecoin reserves — the latest facts free, the history in $2.

Central banks as holders: the ECB’s public-sector purchase programme held €1,837.8 bn at book value on Oct 31, 2026, €245.3 bn less than a year earlier as maturing bonds are not reinvested (Germany €423.3 bn, France €359.5 bn, Italy €278.0 bn); the Bank of Japan held ¥504.9 trn of JGBs on Nov 10, 2026, down ¥50.8 trn (−9.1%) in a year (ECB PSPP breakdown, Bank of Japan Accounts).

Maturity wall by country, and index events — inclusion in WGBI or GBI-EM and the forced flow that follows.

OECD ↗ · FRED ↗ · ECB Data Portal ↗ · BIS debt securities ↗ · US Treasury TIC ↗
Tokenized government debtFive kinds of record, never added together: a new digital government bond, a token on an existing bond, a fund share, a collateral (repo) operation, and bills held in stablecoin reserves. Collateral and cash 24/7: a calculator, not a broker — the net yield of each kind of cash after fees and the collateral haircut, and whether the money can be used over a weekend. Each fund is read in Portfolio Guard (what a product actually pays); Citi’s “Tokenization 2030” forecast — in the Accuracy Ledger. Free — the facts and the method; $2 — the daily net-yield table, open to all after 30 days; One — on your amounts; Desk — on your positions; contract — a feed of your counterparties’ terms
Collateral and cash 24/7: the kinds of cash and the sources

For each kind of cash — a money-market fund, a tokenized bill fund, repo → the net yield after fees and the haircut, weekend availability and, for a tokenized fund, its size on the date from the issuer’s or the chain’s own record; each line with its date and primary source, “planned” where there is none. Sizes are a column of the table, not a ranking of issuers.

Citi GPS ↗
StablecoinsHow many stablecoins are in circulation and what backs them, from the issuers’ own attestations, each figure with its attestation date. Circulation is a stock, so it is never added to flows or to the count of crypto owners. Free — the latest attestations; $2 — series by issuerCircle ↗ · Tether ↗
Crypto741mn owners in 2025Crypto.com ↗
ATLAS · STATE MONEY · ALLOCATION SHIFTS

Allocation shifts of state funds

“Beyond Wall Street” can mean where you buy, where the asset is or how you hold it. A local pension or fund may own US stocks, and local property still moves with global rates — the Atlas keeps those links visible instead of drawing borders. Where a market is hot by sector, it shows on the Country × Sector board.

2027 targets for the S&P 500 span Jefferies’ 9,000, UBS’s 8,900, Barclays’ 8,800, Goldman’s 8,700 (+14%, with earnings growth slowing from 18% to 7%; Yahoo Finance, Nov 13 ↗) and Morgan Stanley’s 8,300 — its horizon is mid-2027, not year-end (the latest set, compiled by Zacks ↗; the next compilation replaces it; each target is now a house forecast in the Journal, with its date and link, and is scored at its horizon against a random picker that draws 5,000 calls on the same day and horizon from the index’s own history). Neuberger Berman favors companies with measurable AI returns over a broad AI bet; Columbia Threadneedle points to semiconductors, Asian EM, local-currency EM debt and commodities, while warning about valuations and concentration; PGIM adds infrastructure debt and parts of private credit.

HOUSE FORECASTS · RANDOM PICKER

Banks’ S&P 500 targets against a random picker

Loading the engine slice…

BANKS’ TARGETS · SEALED IN THE ACCURACY LEDGER

Banks' targets, sealed in the Accuracy Ledger

Loading the engine slice…

Where capital may go in 2027WhyWhat could stop it
AI and its infrastructureDemand for compute and automationWeak payback, debt, overcapacity
Grids and power supplyData centers and electrificationConstruction delays, regulation
Companies lifting profits with AIMeasurable productivity gainsAI gains never reach earnings
High-quality bondsIncome and diversificationInflation and rate hikes
Infrastructure and private creditProjects need financingDefaults, low liquidity
GoldHedging specific risksHigh real yields, a strong dollar
In: a pool of money (ETFs, venture, capex, private credit; other exchanges, household saving routes, pensions, state funds, government bonds of 20 countries) and a period · Does: pulls each figure with its source, date and denominator; never sums unlike pools, never counts accounts as people or turnover as assets · Out: a dated flow map with the forecasters’ range and, for every destination, what could interrupt it — refreshed as new flow reports land.

InvestmentNews ↗ · ETF Trends ↗ · State Street ↗ · iShares ↗ · S&P Global · ETF review Q1 2026 ↗ · ETFGI · 2025 ↗ · ETFGI · Jan 2026 ↗ · ETFGI · Q1 2026 ↗ · McKinsey · Asset management 2026 ↗ · World Gold Council ↗ · Crunchbase ↗ · IEA ↗ · Morningstar ↗ · Neuberger Berman ↗ · Columbia Threadneedle ↗

ATLAS · GOVERNMENT BONDS OF THE WORLD · ENGINE SLICES

Government bonds of the world: the full table, the risk behind the rate, and history

Country map: 10-year yields, slope, real yield, spreads, debt, interest, foreign holders, weights · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16
country10Y, %10Y−2Y, bpreal, %spread to Bund, bpdebt, % GDPinterest, % revenueforeign holders, %weight by debt, %
United States4.99 2026-0924.00 2026-11-122.86 2026-11-12157.82 2026-09124.38 202510.42 202529.26 2026-0839.42 2025
China1.68 2026-09in developmentin development-179.41 2026-0998.61 20253.85 2025in development19.88 2025
Japan3.05 2026-09101.70 2026-11-13in development-40.08 2026-09204.78 20250.55 2025in development9.37 2025
Germany3.47 2026-0928.00 2026-11-13in development—62.66 20251.82 2025in development3.29 2025
France4.44 2026-09in developmentin development96.59 2026-09115.53 20253.98 2025in development4.01 2025
Italy4.36 2026-09in developmentin development93.55 2026-09137.10 20257.51 2025in development3.59 2025
United Kingdom5.24 2026-09in development2.02 2026-11-11183.67 2026-09102.06 20256.82 2025in development4.21 2025
India7.06 2026-09in developmentin development371.05 2026-0984.26 202524.58 2025in development3.38 2025
Brazil9.14 incomparable 2026-0955.00 2026-11-137.77 2026-11-13no data93.33 202518.46 2025in development2.19 2025
Canada3.88 2026-09in developmentin development44.61 2026-09113.58 2025-0.55 2025in development2.71 2025

Rules: Each cell has its date and source. Real and nominal yields are separate columns; spreads only between comparable points of the same month; empty cells are 'plan' or 'no data', never 0. Index weights: the 20 weighted by debt and by GDP — the public rule, not a provider's weights.

Sources: OECD DF_FINMARK IRLT (10-year government bond yields, monthly) ↗ · FRED DGS10 ↗ · FRED DFII10 ↗ · www.imf.org ↗ · US Treasury TIC SLT table 5 ↗ · Ministry of Finance Japan — JGB interest rates (historical + current) ↗ · Deutsche Bundesbank BBSIS ZAR 10y ↗ · Bank of England IUDMNPY ↗ · Bank of England IUDMRZC ↗ · Tesouro Direto — preços e taxas ↗

Yield points on one schema and the recount checks · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16

countries with data 17 · points 35 · shown 10

countryinstrumentdate%annual-compounded %comparability
USAFRED DGS102026-11-125.1005.165comparable
DEUBBSIS ZST R10XX2026-11-133.430incomparable
GBRBoE IUDMNZC2026-11-115.211incomparable
DEUBBSIS ZAR R10XX2026-11-133.4103.410comparable
GBRBoE IUDMNPY2026-11-115.2045.272comparable
JPNMOF JGB 10Y2026-11-132.8992.920comparable

recount check Fed GSW curve: zero-coupon → par: zero -0.0022, par -0.0045 bp, within 1 bp: yes 2026-11-06

recount check Treasury auction: price → yield: 0.0021 bp, within 1 bp: yes 2026-11-10

Daily yields beyond these publishers come from the central banks and clearing houses themselves: on Nov 16, 2026 the 10-year stood at 3.40% for AAA euro-area governments, 3.81% in Canada, 3.11% in Sweden, 4.39% in Norway, 1.63% in China and 8.70% in South Africa (R209, 2036); India’s 6.94% G-sec 2036 yielded 7.26% on Nov 17 and Brazil’s NTN-F 2037 15.21%. Instead of a CDS — the spread on the same day: China −349 bp to the US 10-year, Canada −131 bp, South Africa +358 bp, Brazil +1,009 bp; India +387 bp to the Bund (ECB, Bank of Canada, RBA, Riksbank, Norges Bank, SARB, ChinaBond, CCIL, Tesouro Direto).

Rules: One record per point for any country. Restated to an annual-compounded yield to maturity only where the method allows; a zero-coupon point is not a yield to maturity and stays 'incomparable' unless the full curve gives a par yield; real and nominal yields never share a column.

Sources: FRED DFII10 ↗ · FRED DGS10 ↗ · FRED DGS2 ↗ · FRED DGS20 ↗ · OECD DF_FINMARK IRLT (10-year government bond yields, monthly) ↗ · Ministry of Finance Japan — JGB interest rates (historical + current) ↗ · Deutsche Bundesbank BBSIS ZAR 2y ↗ · Deutsche Bundesbank BBSIS ZAR 10y ↗ · Deutsche Bundesbank BBSIS ZST 2y ↗ · Deutsche Bundesbank BBSIS ZST 10y ↗ · Bank of England IUDMNPY ↗ · Bank of England IUDMNZC ↗ · Bank of England IUDMRZC ↗ · Bank of England IUDSNZC ↗ · Tesouro Direto — preços e taxas ↗ · Federal Reserve Board — GSW nominal yield curve (feds200628) ↗ · TreasuryDirect auctions (Note) ↗

Fiscal-limit monitor (scenario at a threshold of 15% of revenue) · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16
countryinterest / revenue on the slice date, %critical 10Y, %margin, bpflag
United States10.424.76-29.3limit (scenario at this threshold)
China3.853.37169.0attention
Japan0.553.8780.0attention
Germany1.829.85637.1normal
France3.986.14170.0attention
Italy7.515.55114.0attention
United Kingdom6.825.8857.1attention
India24.584.80-238.1limit (scenario at this threshold)
Brazil18.468.56-673.5limit (scenario at this threshold)
Canada-0.5521.281,736.4normal

Rules: Critical 10-year rate at which interest / revenue reaches the threshold within the horizon as old debt is refinanced; margin = critical − current 10Y. A scenario at a given threshold, not a default point.

Sources: www.imf.org ↗ · fiscaldata.treasury.gov ↗ · OECD DF_FINMARK IRLT (10-year government bond yields, monthly) ↗ · Tesouro Direto — preços e taxas ↗

Who holds the debt · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16

TIC 2026-08 · Holdings by country of the custodian, not of the owner (TIC method)

holder (TIC: country of the custodian)USD bn1 month12 months
Japan1,068.3-35.6-115.6
United Kingdom1,009.511.2108.2
China, Mainland590.7-27.3-109.0
Belgium490.519.839.4
Canada464.638.320.6
Cayman Islands463.02.952.5
Luxembourg448.26.137.4
France363.415.0-25.7
Ireland352.11.912.1
Taiwan298.42.3-15.0
Switzerland285.81.0-17.2
Singapore277.90.216.9
Hong Kong261.53.29.6
India219.717.10.3
Norway214.07.1-0.9

central bank: USA: Federal Reserve (H.4.1, Treasury securities held outright) — 4,586.6 USD bn (2026-11-11) · JPN: in development · GBR: in development · DEU: in development

Sources: US Treasury TIC SLT table 5 ↗ · FRED TREAST ↗

Maturity wall · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16

USA · 2026-10-31 · largest year: 2027 — 6,715.8 USD bn

yearUSD bn
20264,791.0
20276,715.8
20283,735.7
20292,664.4
20302,057.5
20312,026.2
20321,200.5
20331,010.5
2034635.0
2035675.7
2036538.9
203738.0

JPN, DEU, GBR, FRA, ITA, CAN, BRA, IND, CHN: in development

Sources: fiscaldata.treasury.gov ↗

Index events · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16
indexcountryannouncedeffectiveforced flow
FTSE World Government Bond Index (WGBI)KOR2024-10-082025-11-01in development
FTSE Emerging Markets Government Bond Index (EMGBI)IND2024-10-082025-09-01in development
Bloomberg EM Local Currency Government IndexIND2024-03-052025-01-31in development

Rules: Forced flow = weight × assets that follow the index; without a primary source for those assets it is not shown as a number.

Sources: FTSE Russell — FTSE World Government Bond Index (WGBI) ↗ · Bloomberg Index Services — Bloomberg EM Local Currency Government Index ↗

Government bond auctions · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16
issuerdatetermbid-to-coverhigh yield, %tail, bptail, σ
USA2026-11-1230-Year2.365.4419.11.62
USA2026-11-1010-Year2.465.02014.02.34
USA2026-11-093-Year2.534.6273.71.08
USA2026-10-297-Year2.464.894-5.6-0.65
USA2026-10-275-Year2.334.8450.50.04
USA2026-10-262-Year2.594.7081.80.65
USA2026-10-2120-Year2.735.469-0.1-0.07
USA2026-10-0830-Year2.545.618-5.2-1.15
USA2026-10-0710-Year2.775.3003.00.28
USA2026-10-063-Year2.624.932-3.8-1.59

Next auctions

issuerdatetermUSD
USA2026-11-1820-Year16.00 bn
USA2026-11-1910-Year19.00 bn
USA2026-11-232-Year
USA2026-11-241-Year 11-Month
USA2026-11-245-Year
USA2026-11-257-Year

history: 500 auctions, 29 bond-vigilante flags

other issuers (DEU, GBR, JPN, ITA, FRA, CAN, BRA, IND, CHN): in development

Rules: Tail = high yield − pre-auction yield; no pre-auction yield → no tail. Surprise in σ against the issuer's own last 24 auctions of the same term. The vigilante flag needs all three: weak auction, term premium up and currency down in the same week.

Sources: TreasuryDirect auctions (Note) ↗ · TreasuryDirect auctions (Bond) ↗

Bond history: what happened after a yield crossed a level · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16

eras: Nov 2011 · Italy 10-year near 7% · May 2013 · eve of the taper tantrum · Sep 2022 · the day before the UK mini-budget (gilt stress) · Dec 2022 · BoJ widens the yield-curve band · Jul 2023 · BoJ makes YCC 'flexible' · Oct 2023 · BoJ: 1% becomes a reference · Mar 2024 · BoJ ends YCC and negative rates

Japan 10-year government bond yield (daily) above 1.0% — what happened next

crossed onvalue, %+3 months+6 months+12 months
2003-07-031.0961.3831.3631.785
2010-08-271.0291.1871.2551.066
2011-09-081.0111.0120.9840.811
2012-03-141.0040.8500.7940.615
2024-05-221.0070.9191.0921.559

United Kingdom 10-year government bond yield (daily) above 4.5% — what happened next

crossed onvalue, %+3 months+6 months+12 months
2022-09-274.5273.8203.4714.566
2023-06-194.5274.2713.6204.078
2024-11-054.5534.4484.4524.467
2025-05-084.5384.5894.4874.870
2025-11-144.5734.4175.126

Italy 10-year government bond yield (OECD, monthly average) above 4.0% — what happened next

crossed onvalue, %+3 months+6 months+12 months
2006-12-014.0384.1764.7724.538
2010-06-014.1023.8624.6034.816
2013-06-014.3794.5414.1142.920
2022-09-014.1384.2574.2404.513
2026-09-014.362

United States 10-year government bond yield (OECD, monthly average) above 4.5% — what happened next

crossed onvalue, %+3 months+6 months+12 months
2005-11-014.5404.5705.1104.600
2023-10-014.8004.0604.5404.100
2024-04-014.5404.2504.1004.280
2025-01-014.6304.2804.3904.210
2026-07-014.600

Rules: base = last value known at T 09:00 ET; h-month result = last value whose known_at ≤ (T + h months) 09:00 ET

Sources: OECD DF_FINMARK IRLT (10-year government bond yields, monthly) ↗ · FRED DGS10 ↗ · FRED DGS2 ↗ · FRED DGS3 ↗ · FRED DGS5 ↗ · FRED DGS7 ↗ · FRED DGS20 ↗ · FRED DGS30 ↗ · FRED DFII10 ↗ · FRED TREAST ↗ · FRED THREEFYTP10 ↗ · FRED DTWEXBGS ↗ · FRED SOFR ↗ · FRED DTB3 ↗ · FRED DEXBZUS ↗ · FRED DEXINUS ↗ · FRED DEXUSEU ↗ · FRED DEXUSUK ↗ · FRED DEXJPUS ↗ · Deutsche Bundesbank BBSIS ZST 2y ↗

ATLAS · GOVERNMENT BOND AUCTIONS · GERMANY AND JAPAN

Government bond auctions outside the US

ATLAS · GOVERNMENT BOND AUCTIONS · UNITED KINGDOM

UK gilt auctions

ATLAS · CASH 24/7 · TOKENIZED DEBT · STABLECOINS

Cash 24/7, tokenized government debt and stablecoins

Tokenized government debt: five kinds of record, never added · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16
recordkindissuerdatesize
HKSAR-DIGITAL-GREEN-2024 ↗new digital government bondHKSAR Government2024-02-076.00 bn HKD
BUIDL-ETHEREUM ↗fund shareBlackRock USD Institutional Digital Liquidity Fund (BUIDL), Securitize transfer agent2026-11-16170.28 mn USD
TOKEN-ON-EXISTING-BONDtoken on an existing bondno issuer record with a primary source wired yetin development
TOKENIZED-REPOcollateral (repo) operationrepo platforms (e.g. intraday DLT repo) publish monthly volumes in press releases; parser planin development
STABLECOIN-RESERVE-BILLSbills in stablecoin reservesTreasury bills in stablecoin reserves come from issuers' attestations (snapshot stablecoins, §35)see stablecoins

Rules: The five kinds are different records: never summed with each other, with stablecoins or with money-market funds; a wrapper is not a second count of the same bills.

Citi «Tokenization 2030»: in development a complete open counter of tokenized assets is not wired; the measured part is not divided by the 5.5 trillion target

Sources: HKSAR Government — HKSAR-DIGITAL-GREEN-2024 ↗ · Ethereum eth_call totalSupply 0x7712c34205737192402172409a8F7ccef8aA2AEc ↗

Cash 24/7: net yield after fees and haircut · slice as of 2026-11-16
Engine slice · slice as of 2026-11-16
kind of cashyield, %fees, %haircutweekendnet, %
3-month Treasury bill held directly (TreasuryDirect)4.140.000.00no4.14
Overnight Treasury repo (SOFR)3.90in developmentin developmentnoin development
Tokenized Treasury money-market fund (BUIDL)in developmentin developmentin developmentyesin development
Government money-market fundin developmentin development0.00noin development

net = yield − fees − haircut × funding cost

Rules: A row without a primary source for its yield or haircut is 'plan', not a number. A calculation, not a brokerage function: no orders, no advice on releasing collateral.

Sources: FRED DTB3 ↗ · www.treasurydirect.gov ↗ · FRED SOFR ↗

Stablecoins: circulation and reserves from the issuers · slice as of 2026-11-15
Engine slice · slice as of 2026-11-15
issuercoinin circulationreserves (total assets)attestation
TetherUSDT175.40 bn USD180.86 bndate not in the issuer’s feed
TetherXAUT764,235 troy_ounce827,418date not in the issuer’s feed
TetherMXNT19.56 mn MXN19.56 mndate not in the issuer’s feed
CircleEURC472.01 mn EURmonthly report, date on the linked page
CircleUSDC69.07 bn USDmonthly report, date on the linked page

flows: accumulating in development

Rules: issuer's own publication only · stock (circulation, reserves) and flow (net issuance) are not added

Sources: tether.to ↗ · Tether transparency ↗ · www.circle.com ↗ · Circle stablecoins API ↗

3.2 — SCENARIO LAB

Spot the regime shift — don’t guess tomorrow’s price

Choose an asset. See the bull case, the bear case, the arbiter’s read, the model votes and three conditional scenarios. Facts and votes come from the council’s answer of Nov 14, 2026; probabilities come from the live council.

SPOTTED AN OPPORTUNITY? CHECK YOUR IDEA

From “I have a hunch” to checked reasons to act: one card turns a guess into fields you can test and names the one unknown that would change your decision and the most useful way to check it.

    Freeone asset in full, the conclusions for the rest$2every asset: debate, votes, three scenarios, the AI-capex meter, the archive, the bubble-test status of every assetOneyour own asset
    CAPEX METER · PAYBACK CHECKS · COMPANY BY COMPANY

    Does AI spending pay back? Three checks per company

    the three payback checks, company by company

    AI-revenue payback: three checks per company · T = 2026-11-15

    US

    CompanyStandard1 · Who pays2 · What remains3 · Will it repeat
    MSFT US GAAPnot disclosedpassedpassed
    NVDA US GAAPnot disclosedpassednot indicative
    ORCL US GAAPnot disclosedpassedpassed
    AMZN US GAAPnot disclosedpassednot disclosed
    GOOGL US GAAPnot disclosedpassedpassed
    META US GAAPnot disclosednot passednot disclosed

    Asia

    CompanyStandard1 · Who pays2 · What remains3 · Will it repeat
    TSM TSMCIFRSnot disclosednot disclosednot disclosed
    BABA AlibabaUS GAAPnot disclosednot passednot disclosed
    005930.KS Samsung Electronics———
    000660.KS SK hynix———
    0700.HK Tencent———
    9984.T SoftBank Group———

    no sums across companies or columns: one dollar would be counted at the seller and the buyer, and different kinds of revenue would be added; no payback date is produced. A person checks 5 results before they count: waiting for a person in development

    Engine slice v43 ai_payback · as of 2026-11-15 22:20:57 UTC · sources: SEC 10-K 0001564590-20-034944 ↗ · SEC 10-K 0001564590-21-039151 ↗ · SEC 10-K 0001564590-22-026876 ↗ · SEC 10-K 0000950170-23-035122 ↗ · SEC 10-K 0000950170-24-087843 ↗ · SEC 10-K 0001193125-26-323660 ↗ · SEC companyfacts MSFT ↗ · SEC 10-K 0001045810-26-000021 ↗ · +29

    More: bubble test — when a run turns explosive
    LAB · BUBBLE TEST · PSY 2015

    Bubble test — when a run turns explosive

    Gold, copper, the Nasdaq, bitcoin and US house prices (also against rents): a recursive explosive-root test (Phillips, Shi & Yu, 2015 ↗) dated observation by observation. It dates the start and end of a run; it sets no price target.

    Gold: explosive phase since Apr 2024 (slice of Nov 17, 2026; October 2026 average $4,105 after the February record of $5,020).

    The bubble test now covers the S&P 500 since 1871 on Robert Shiller’s monthly series: the price/dividend ratio has no open phase (the last ended in May 2009; the test also dates 1997–2001), the real price has been in an explosive phase since June 2024, CAPE 39.4 — above 98% of months since 1881. Across the 12 Lab markets and 10 board markets (weekly since 2002), two are in an explosive phase today: the PHLX Semiconductor index since Jun 27, 2025 and Taiwan’s TAIEX since Nov 28, 2025 (Shiller data; PSY 2015).

    Bubble test: an asset’s price — or price against earnings, rent or the cost of production — on any exchange → a recursive explosive-root test, dated day by day (the method of Phillips, Shi & Yu, 2015 ↗) → “explosive phase since [date],” “no explosive phase” or “the phase ended on [date],” with the series and its critical value. It dates the start and end of a run; it sets no price target. The first worked case is gold around its January record and the fall since (Gold tab); the same test reads an index, a single stock, copper, bitcoin or a housing index. A new phase becomes a Thesis Monitor event and a Portfolio Guard stress. Free — a worked gold example and the method; $2 — the current status of every Lab asset and every market on the board, open to all after 30 days; One — your asset.
    council debate
    Bull case▲

    Skeptic▼

    Arbiter⚖

    COUNCIL VOTES · FROM THE NOV 14, 2026 POLL
    Base case
    Upside case
    Downside case
    3.3 — BOARD AND RADAR

    Where money is hot — and where the odds split

    Pick a market and a sector on the board, then compare council probabilities with prediction-market prices.

    LIVE · HEAT MAP · MARKET × MODEL

    Heat map: market × model

    Colour is the side (green — up, red — down), strength is the confidence. Pick the horizon.

    Loading… Trends ↗

    0%100%
    LIVE · SEALED BEFORE THE OPEN · 12 MONEY × 4 HORIZONS

    12 markets, 4 horizons: what the AIs would do

    Sealed before the New York open. Pick a market: the Council’s call, every model’s target, entry and “wrong if”, the 30-session chart with those levels, and the crowd’s price.

    Loading the day’s sealed answers… Open on Trends ↗

    Save to my thesis →

    COUNTRY × SECTOR BOARD · PICK A CELL · SCORED DAILY

    Where is capital flowing — which market, which sector?

    Pick any cell — such as A7, Taiwan × chips — or narrow the board by market and sector: you get its coordinate, the dated fact behind it when there is one, the local exchange and a ready question for the council. Views: board, by market, by sector, map, flows.

    Freethe grid, any cell, coordinates, facts, views, the question to the council$2cell scores, the market / sector / rest split, the trail and rotation, scored dailyContractdata export
    Market ↓ · sector →A · Chips & hardwareB · Software & internetC · Energy & powerD · FinancialsE · Industrials & roboticsF · Materials & metals
    1 · United States
    2 · China (mainland)
    Stock Connect carries about 71% of international investors’ holdings of mainland A-shares — around RMB 2.6tn at end-2025. Hong Kong is the gate between China and outside money. HKEX ↗
    4 · Japan
    5 · India
    Korean retail investors bought $7.63bn more US stocks than they sold in Q3 2026: local savings heading abroad — a cross-border flow the board follows (see the Flows view). Korea Times ↗
    7 · TaiwanSemiconductors were 52.22% of Taiwan Stock Exchange market capitalisation at end-2025: 94 companies, NT$49.28tn. Here a study of the chip supply chain says more than a view on “Asian equities”. TWSE Fact Book 2026 ↗
    8 · Europe
    Since Feb 1, 2026 the main market is open to direct investment by all foreign investors, individuals included, with no QFI status required (CMA, announced Jan 6). A change in access is itself an event the board tracks. CMA ↗ · Regulation Tomorrow ↗
    10 · Brazil
    CELL DATA · USD MOVE OVER 5 SESSIONS · WHAT DROVE IT

    Every cell with its numbers: the week’s move, what drove it, momentum and rank

    Loading the board slice…

    ROTATION · 10 MONEY AND 6 SECTORS AGAINST MSCI ACWI · WEEKLY

    Loading the rotation…

    • 6 → 1 · South Korea → United States. Korean retail investors bought $7.63bn more US stocks than they sold in Q3 2026 (a flow over one quarter). Korea Times ↗
    • 3 → 2 · Hong Kong → China (mainland). International investors held around RMB 2.6tn of mainland shares through Stock Connect at end-2025, about 71% of their total — a holding at a date, not a flow, so it is never added to one. HKEX ↗

    Marked cells and rows carry a dated, sourced fact; every other cell can still be picked, filtered and taken to the council. Colour by strength comes from the engine’s daily score, the flows from each market’s fund files, the rotation from weekly relative strength against MSCI ACWI; no view draws a made-up score.

    Method: what goes in, how the score works, what comes out

    Investors read two planes at once: where (the market) and what (the sector). The board puts both on one grid, like coordinates in Battleship: a hot spot is a cell such as A7 — Taiwan × chips. Every view draws the same dated snapshot, and rotation joins them with the engine’s scores.

    In: date T, up to 10 markets and 6 sectors, and what “country” means — listing, domicile or revenue (MSCI ↗ shows the three draw different maps; switch the definition in the block below — listing and domicile are drawn today on the same dated snapshot; revenue — from each company's own annual report (SEC XBRL), for the largest listed companies) · Does: for each cell measures momentum against MSCI ACWI, fund flows and earnings revisions (with a licensed estimates feed), then splits the cell’s move into a market effect, a sector effect and the rest (Heston–Rouwenhorst regression) · Status: working today — any cell, market and sector filters, five views with data on one dated snapshot (board, by market, by sector, map, flows) plus rotation, the local exchange and a ready question for the council; your view, filters and cell are remembered. Plan — the daily score; the first scored column is chips across eight markets (all on the board except Saudi Arabia and Brazil), the AI-infrastructure route already in the plan · Out today: the coordinate, its dated fact, the local exchange and a prepared question in the council form. Out with the score: the cell of the day with its reason (“mostly sector, partly market”), the week’s trail and what switched — market, sector or both; the cell is sealed in the Accuracy Ledger and scored after a week and a month; from a cell, its chain in the Sector Researcher and a Follow button that makes it a thesis in the Thesis Monitor.
    COUNTRY × SECTOR BOARD · WHAT “COUNTRY” MEANS

    What “country” means on the board

    Loading the board…

    DIVERGENCE RADAR

    Where AI odds and market odds split

    Compare council probabilities with prediction-market prices and futures. Kalshi and Polymarket reached a record $50.6bn in July 2026 (Tangem ↗), and Kalshi a $15.27bn record week in September (DeFi Rate ↗). A wide gap flags an idea to investigate, with both sides laid out; after the event, the Accuracy Ledger shows who was closer — the crowd with money at stake or the council.

    Live on 12 markets: on 49 crowd contracts the AIs and the crowd take the same side 92% of the time, but the crowd sits 22.7 points from 50 and the AIs 8.9: money is bolder. Each day the widest gaps lead (the Spread, the Crowdometer), and the Polymarket–Kalshi pair shows one market at two prices. The free divergence of the day: the dollar index — the crowd 15.6 points pricier than the Council (live page, Nov 16, 2026). Cut-off: session of Nov 13, 2026, data Nov 16, 2026, 11:38 ET, export ↗.

    LIVE · THE CROWD NEXT TO THE COUNCIL

    Loading today’s crowd prices… Crowdometer on Trends ↗

    More: beyond the US, like-for-like contracts, one asset at two prices
    Beyond the US: the same comparison against expectation surveys — Brazil’s Focus survey (Banco Central do Brasil) and the ECB Survey of Monetary Analysts: the median expected rate for the next meeting against the rate in force, and after each meeting the surprise. And against the futures themselves, from the exchanges’ own settlement prices: Eurex €STR futures (Nov 16) price €STR at 2.68% after the Dec 23 ECB meeting and 3.16% by the end of 2027, +72 bp against today’s 2.44%; B3 DI1 rates imply a forward CDI of 15.27% around the end of 2027, +162 bp against today’s 13.65%. Only the derived rate is shown, not the exchange prices (Eurex, B3, ECB, Banco Central do Brasil).
    BEYOND THE US · ECB SURVEY OF MONETARY ANALYSTS VS THE ECB

    Loading the latest survey…

    Like-for-like contracts: before any comparison the council’s question and the market contract are matched on settlement rule, outcome source, date and spread → a “comparable / partly / not comparable” tag; only comparable pairs enter the divergence and the Brier score. Free — the method and the tag on every pair; partly comparable pairs are shown for reference only.
    LIKE-FOR-LIKE · COUNCIL QUESTION VS CROWD CONTRACT

    Loading the comparability check…

    One asset, two prices: the same company or basket quoted in two places — a US ADR and the home share, Hong Kong H shares and mainland A shares, an ETF and its net asset value, a closed-end fund and its net asset value, the same coin on exchanges that trade it in different currencies (exchange public data) → the premium or discount after currency, its history and the hours when one market is shut → a gap wider than its own history is flagged as a question, with what usually closes it (conversion rules, capital controls, creation and redemption, withdrawals between exchanges). Closed-end funds are tracked daily: 19 funds (PIMCO, Nuveen, Eaton Vance, Gabelli, Cohen & Steers, BlackRock and others) with a same-day price and NAV — median premium −7.2% on Nov 16, 2026; 11 sit outside their own one-year range (Nuveen NEA −6.1% against a usual −3.4% … +1.1%; Eaton Vance EOS −14.3%). Free — one pair a day; $2 — the full list, open to all after 30 days.
    ONE ASSET, TWO PRICES · PREMIUM VS ITS OWN YEAR

    Loading today’s pairs…

    CLOSED-END FUNDS · PRICE VS NAV

    Closed-end funds: price against net asset value

    Loading the engine slice…

    Freeone divergence of the day; the top three with a delay$2the full list of the day and each divergence’s historyProyour own markets, futures series beyond the US
    LIVE · MAP +100%

    Map +100% · where the AIs saw a doubling

    Sealed before the open: each AI could name one market that may double within a week. Open on Trends ↗

    Doubling in a week is rare. These are the AIs' own guesses; we score them when the week closes, and misses are shown too.

    LIVE · US MODELS × CHINESE MODELS

    East against West

    The same question, the US models against the Chinese ones: the mean chance of a rise per group, market by market.

    LIVE · EVERY DAY · TOP-10 WITH REASONS

    Ratings of the day

    Today’s top lists load from Trends. Ratings of the day ↗

    3.4 — HISTORY ENGINE

    “10-year above 5%.” What happened next?

    Money finds past episodes, shows what followed and tests which similarities still matter.

    COMPACT BREAKDOWN · WHAT WAS KNOWN THAT MORNING

    History Engine: how an era is like today and how it differs

    Five eras, each given as it looked on the date T. The outcome and the lesson stay closed until you press Reveal, because the name is hindsight.

    EpisodeBackdrop on date T10-year after 12 mo.Gold after 12 mo.Copper after 12 mo.Similar to todayNot similar
    Jun 1, 2007 · 10-year near 5%Fed funds at 5.25% for a year; several subprime mortgage lenders have collapsed since FebruaryHigh rates, debt risingThen the debt sat with households and mortgage lenders; now it sits with AI companies
    Oct 18, 2023 · 10-year near 5%Fed funds at 5.25–5.50% after 11 hikes; AI capex risingAI capex, concentration of leadersThen the Fed had paused after 11 hikes; now the market expects a hike
    Sep–Nov 2026 · 5.1–5.2%Hot inflation, capex $0.7–0.8tn, debt for AI→ distribution across analogs + the Council 10+ assessmentThe council explains which episode is closer and why
    historical regimes compared with November 2026

    Similar to today
    Not similar

    replay · what was known on date T

    Pick a date. See the market as it looked that morning

    Freea demo episode, the method, the chart up to T$2all 63 dates with Reveal, and the archiveProyour own strategy, backtested

    Hindsight makes every crash look obvious. Replay rebuilds a past morning from what was public by 09:00 ET on date T: the 10-year yield as published (it is never revised), payrolls as first reported (ALFRED), and copper and gold cut at T by their publication lag. Press “Reveal” to load what happened next.

    In: a date T and an asset (the 10-year yield, copper or gold) · Uses: FRED and the World Bank Pink Sheet with a publication time for every point, ALFRED vintages for payrolls · Out: the chart as it looked on the morning of T, then — on “Reveal” — the path that followed and the change after 3, 6 and 12 months, computed by the engine. Next: SEC filings by acceptance time, index members on T and the council’s sealed call beside the outcome; then local series beyond the US — each central bank’s rates with their release dates.

    Both episodes start with the 10-year near 5%. Press “Reveal in replay” to see what followed each one; the figures come from the engine’s after-T snapshot. The replay has 63 prepared dates since 1971 for the 10-year, copper and gold; oil, the Dow and currencies are next. Government bonds of every country: the same question on the 20 bond markets of the Atlas — “Japan’s 30-year above X: what happened next?” — with the 2022 gilt stress and the Bank of Japan’s exits from yield-curve control among the episodes. Free — a demo; $2 — every date.

    How Money asks history an honest question — and the backtest safeguards
    ask history a precise question

    “Using only what was known on date T, can we improve the risk or return estimate for asset X over horizon H?”

    “Feed the model the whole internet” is not a research design. Qlib and LEAN already exist to test models and strategies; Money competes on better data, a sharper question and a stricter test. History sets ranges, scenarios and risk. Forecasts are locked before the outcome and scored on genuinely new data — a model can earn its keep by managing risk better even when it can’t call tomorrow’s price.

    MarketHypotheses Money tests
    GoldReal yields, inflation expectations, the dollar, ETF and central-bank flows
    CopperIndustrial demand and Chinese construction, inventories, mine disruptions and new capacity, grid build-out
    Equity indexEarnings and margins, valuations, rates and credit spreads, buybacks, market breadth, index composition and concentration
    safeguards built into every backtest
    • Overfitting → walk-forward testing and an untouched sample.
    • Look-ahead → only data published by date T, using ALFRED vintages for macro series.
    • Survivorship → index members as they stood on each date, including delisted companies.
    • Costs → fees, bid-ask spreads and liquidity in every result — plus each exchange’s own rules from the Market Profile: transaction taxes such as India’s securities transaction tax, lot sizes, daily price limits and settlement (trading days: exchange_calendars ↗), so a trade that could not have been executed is rejected. Free — the exchange-rules reference; Pro — applied to your backtest. Worked today on six listings — Shanghai, Taipei, KOSPI, NSE, Tokyo and Tadawul — with one rule (hold the stock above its 200-day average, otherwise cash): over ten years the transaction tax alone took 10.6 points from TSMC’s result (+496% frictionless, +484% by the exchange’s rules) and 2.1 points from Reliance’s; a sell order on Tokyo’s Greenery Day 2018 was not filled and waited for the next session.
      EXCHANGE RULES IN THE BACKTEST · SIX MONEY

      Loading the exchange-rules backtest…

    • Regime change → results reported by regime, not as one average — local regimes too: Japan 1989–90, Asia 1997–98, UK gilts 2022, Brazil 2015, with cash at the local central-bank rate and the result in the currency you spend. Regime comparison: an asset, a spending currency and a horizon → regimes set from inflation, real yield, currency and volatility as known on date T (a hidden Markov model, statsmodels) → the range of outcomes for that asset and for cash in past matching regimes, how many matches and which dates, and the condition that breaks the analogy; fewer than 10 matches → “regime outside the analogs,” with no probability. It feeds Portfolio Guard and the Lab. Free — the method and one dated example; $2 — the shared regime map while fresh, open after 30 days; Pro — your strategy by regime. Worked today: on September 2026 data the four drivers read high inflation, a high real yield, a weakening dollar and calm markets; over three months that regime has 12 independent past matches — gold’s median +0.6% (10–90%: −5.9% to +15.6%) against cash +1.0% — while at six and twelve months there are fewer than 10, so the answer is “outside the analogs”, with no range. Scores that waited for calibration are now published, each with its interval and its distance from a benchmark (walk-forward 1973–2022, Brier score against the base rate): a US recession within 12 months — 27% (90% interval 18–38%, Oct 2026), +0.054 ±0.026 better than the benchmark (skill +0.27); a negative S&P 500 real total return over 12 months — 31% (25–37%), −0.002 ±0.002 against the benchmark: the regime model does not beat the base rate yet, and the label says so (FRED, NBER dates, Shiller).
      REGIME COMPARISON · FOUR DRIVERS ON DATE T

      Loading the regime comparison…

    • Multiple testing → next in the engine: every variant tried is logged, the Sharpe ratio is deflated for the number of trials, and a single test across all variants at once (White’s Reality Check) runs across all variants — a thousand generated strategies will always contain a lucky one.
    • Model memory → a frontier LLM may already know how 2020 ended; telling it “ignore the future” isn’t enough. Money grades LLMs only on forecasts made after their training cutoff, an issue the Federal Reserve and research on time-restricted models both study. Gao, Jiang and Yan built a test for it: in their headline application, a one-standard-deviation rise in a news item’s lookahead propensity strengthens the LLM signal by about 32% of its standalone effect — a sign the model partly remembers rather than forecasts; it holds for that test, not for every model (v2, June 2026, arXiv ↗). Our own measurement: the same fact pack up to date T runs twice — with the real names and dates and anonymized; the gap between the answers is each model version’s “memory gap,” and a backtest with a significant gap stays out of the Accuracy Ledger. Free — the method and the verdict on every backtest; the gap matrix by model stays internal (it feeds the Router). On Nov 17, 2026: 10 model versions from 4 families measured on one pack of 108 items; 5 answer significantly better with real names on dates they may remember, and their backtests over those dates stay out of the Ledger.

      On Nov 17, 2026: 10 model versions measured on the same pack of 108 items, 12 markets — 5 answer significantly better with real names on dates they may remember.

    Why the bar is this high. An April 2026 NBER working paper followed portfolios picked by ChatGPT, Claude, Gemini and Grok: raw returns beat the S&P 500, but the abnormal return after characteristic adjustment was not statistically significant — the picks leaned on momentum, large companies and media attention (NBER w35153 ↗). AMCIS 2026 found Gemini Deep Research no better than a coin flip on Nasdaq-100 stock direction. In Gu, Kelly and Xiu, the best model explained only ~0.4% of an individual stock’s monthly return. In October–December 2025 live-money contests such as Alpha Arena, most models lost money, overtraded and gave inconsistent calls on identical data. Open time-series models — Kronos, trained on 12+ billion price bars, plus TimesFM, Chronos-2 and Moirai — beat traditional methods on several tasks (dated research, 2025–2026). Money’s bar is the statsforecast baseline (naive, ETS); its challenger — the mean of AutoARIMA, AutoTheta and AutoETS — is sealed in the Ledger beside it and scored the same way: on Nov 17, 2026 it beat Naive in 5 of 9 series-horizons (MASE). See What’s inside.

    Challenger (mean of AutoARIMA, AutoTheta, AutoETS) against the statsforecast baseline, MASE on the same origins.

    Free series to start from: FRED copper since 1992 ↗ · World Bank Pink Sheet (gold) ↗ · FRED Dow ↗

    QuantConnect · research guide ↗ · Federal Reserve · FEDS 2025-044 ↗ · Qlib ↗ · ThinkAdvisor ↗ · TSFM in finance ↗ · Chronos-2 ↗

    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.