The extended Watchtower guide

Chapter 01Using this guide #

If you are new, follow Your first fifteen minutes, then Navigation and workspace controls and Reading data correctly. If you already use Watchtower, begin with What is new or search this guide for a workspace, task or term. The final workspace directory accounts for all 78 registered navigation tabs in the audited application; the chapters cover their major subviews and related account, research and developer features.

Use the chapter contents to move through the web edition. Search matches chapter text as well as titles. Clear the search to restore the complete book. The PDF is intended for offline reading and printing; the Markdown download is an editable copy. Neither download refreshes its observations after you save it.

Edition boundary. Interface names and feature availability were checked against the deployed application and published methods on 16 October 2026 UTC. Counts in this guide describe a registry, an explicit example or a dated release snapshot. Current record counts, source conditions, prices, entitlements and model results can change. The panel's own source timestamp and access message govern the view you are using.

Access. The public Daily Summary is available without an account. Pro provides the broader research workspace according to the account's current capabilities. Some reports, AI requests and optional provider connections have separate access or cost conditions. Administrator and reviewer controls are role-gated. Reading this guide does not grant those capabilities.

Useful companions are the data catalogue, release updates, AI-agent connection guide and the feature-specific methods linked throughout this book.

Chapter 02Your first fifteen minutes #

Minutes 1–3: orient yourself

1. Open Watchtower. Start in Daily Summary if you want the public overview, or sign in and use Simple View for a compact research layout. If the sign-in prompt appears, sign in/create an account as appropriate; Not now closes it when you only want the public overview.
2. Open the User & Settings menu. Check Account / Access so you know which account and capabilities are active. Choose a comfortable light or dark theme.
3. Read one market observation and one event. Identify the original source, observation or publication date, units and geographic scope before interpreting the headline.
4. Notice the top status indicator. It describes application/source operation, not a global danger score.

Minutes 4–7: investigate one question

Choose a concrete question: “What changed in Japanese markets?”, “Which warnings cover this country?”, or “What evidence supports this banking concern?” Open the relevant workspace using Help, the tab strip or the Command Palette. Set the country, region, date window or instrument explicitly.

Open an evidence detail, source link or Raw data & source evidence disclosure. Record one fact and one limitation. For example, a port activity series may be dated and regional; a country's annual debt measure may not describe today's borrowing; an official warning describes the issuing authority's product and validity period.

Minutes 8–11: save a useful view

If you have the relevant access, open My Watchtower and create a watch for the country, domain, company, sector, topic or asset you care about. Start with a narrow watch. Read What changed, then configure the briefing time and inbox settings deliberately.

For investigation notes, use the workspace's own save/export controls. Check whether the save is cloud, browser-local or session-only. Copying a view link shares navigation context; it does not transfer private notes, paid data or another person's access rights.

Minutes 12–15: finish with a research record

Export a small CSV or evidence JSON, or save an investigation/report where offered. Write a one-paragraph conclusion containing: the question, source-dated observations, your interpretation, important gaps, and the next condition you will check. Use the end-to-end playbooks later in this guide to expand this routine.

Chapter 03Navigation and workspace controls #

Modes, navigation and Help

Daily Summary, Simple View and Advanced are available from User & Settings, subject to access. Brief, Monitor and Investigate are operating modes in the command navigation. Simple View is useful for Pro Brief, Signal Board, Alert Screen, Search and personal monitoring. Advanced exposes the larger set of specialist workspaces. A layout change can move a tab without removing its data.

The left command navigation includes workspace actions such as Map Layers, Evidence, Timeline, Commands and Tour. Use Hide menu to recover horizontal space. On a phone, some navigation is presented in compact menus or a bottom rail, and tab strips can scroll horizontally.

Help is a panel finder with topic searches and shortcuts. It also offers the Advanced Tour and Deep-Dive Tour. Use the extended-guide link there to return to this handbook. The Command Palette is the faster route when you know the destination name. Research/Finance/OSINT tools may be inside the expandable tools area rather than the primary strip; Help and the workspace directory are useful when a tab is not immediately visible.

Keyboard shortcuts in the application include Ctrl+K / Cmd+K for the Command Palette and / when you are not typing. Left/right arrows move between supported focused tab controls. Escape closes the applicable menu/dialog layer. This handbook's own / shortcut focuses its guide search; browser Find remains available.

Customize, expand and share

• Customize Workspace manages the arrangement and visibility of workspaces. Panel Move controls rearrange panels; a moved panel is not a new data source.
• Full screen expands a workspace. Many maps move the same canvas and controls into the enlarged view. Close the expanded view or use its supported Escape action to return. A nested dialog or layer menu may close before the surrounding fullscreen view.
• Copy view link preserves the supported navigation context. Use the generated link rather than inventing query parameters. Recipients still need their own entitlement and source access.
• Pin in workspace, favorites and Startup Presets help you build a repeatable routine. These settings are distinct from a dataset's own watchlist.
• Wide evidence tables may scroll inside their own container. On compact screens, open the map's Filters or Data rail when you need controls hidden to preserve map space.

Main globe and independent maps

The main 3D globe can start closed. Use Open 3D globe when you need it; close it to reclaim space and stop its own rendering and live-layer activity. Dedicated workspaces such as Aviation, War Room, Commodity Atlas and country atlases retain their own maps and source scopes. The aircraft layer on the main map and the dedicated Aviation tracker are separate views; their counts need not match.

Shared map controls

Where a Layers menu is offered, choose analytical, street, satellite, terrain or topographic backgrounds and adjust reference markers, boundaries, shading or opacity. A satellite basemap is imagery, often a mosaic acquired on different dates. A terrain-style background is not automatically a new elevation analysis. A country colour can represent a statistic, model, evidence count or coverage state: read the selected measure and legend.

Chapter 04Reading data correctly #

Watchtower is most useful when you distinguish what a number actually represents. Open source details before combining figures from different panels.

The five clocks to check

Occurrence/observation time is when the event or measurement belongs. Publication/issue time is when the source released it. Reference period identifies a month, quarter, year or survey interval. Collection/check time is when Watchtower obtained or checked it. Model valid time/horizon says when a forecast applies.

These clocks can legitimately differ. A newly downloaded annual statistic can be several years old. A daily market bar's timestamp is not necessarily its last trade. A warning can be issued now but effective later. A date-only source should not be interpreted as a precise midnight observation.

Missing, zero, stale and partial

A dash, null or unavailable field means there is no qualified displayed value. A genuine zero is a measurement and should remain zero. Stale, retained, aged, partial, reference-only, not integrated and credential required describe different conditions. Read the panel's explanation rather than treating them all as “nothing happened.”

Counts are frequently heterogeneous: articles, CAP messages, language versions, database rows, providers and locations are different units. Do not add them into a total of independent incidents. Duplicate coverage from related publishers or overlapping datasets is not independent corroboration.

Geography and units

A point may be an observed event location, an approximate locality, a reference headquarters, a municipality centre or an aggregate cell. Unlocated evidence can remain in a table without a map marker. A warning polygon is an issuing product's area, not necessarily an observed damage footprint. A model layer carries its own assumptions.

Check nominal versus real values; market USD versus purchasing-power-parity units; stock versus flow; percent versus percentage points; level versus return; quarterly versus year-to-date; and issuer/listing country versus asset exposure. Colour and animation help navigation but do not change these definitions.

Raw evidence and reproducible reading

Raw data & source evidence is normally collapsed at the bottom of the relevant tab. Expand it for official points, domain support, original records and source coverage. Use row information buttons for specific observations. Save source links, original dates, selected filters and model/definition versions with an export. When a snapshot changes during a multi-page export, use the workspace's fresh-read/retry path rather than joining different generations yourself.

Chapter 05Daily Summary, Pro Brief and Signal Board #

These are different entry points for different jobs.

Read the lead items, then open their canonical workspace. A concise card is a pointer to evidence, not a substitute for all source qualifications. When a source is unavailable, follow the coverage detail rather than interpreting an empty card as a normal condition.

Signal Board lets you assemble a focused monitoring display. Search for a signal, pin it, reorder or remove tiles, and expand the board when useful. Source-anchored debt clocks can also be pinned here. Their playback state is shared with the corresponding debt-clock controls during the session; a running animation does not turn a reported stock into a live measurement.

Alert Screen is the compact place to review monitoring conditions. It is not identical to the Alerts rule editor or My Watchtower's server inbox. Use the alert's own origin, threshold, timestamp and evidence to decide where to investigate next.

Chapter 06My Watchtower #

My Watchtower is the personal daily-workflow workspace. It combines cloud watchlists, acknowledged changes, scheduled briefings, an in-app inbox and usefulness feedback. Its settings and results belong to the verified account.

Create your first watch

1. Open My Watchtower and create a clearly named watch, such as “Canada: banking and household debt.”
2. Choose the supported country, domain, company, sector, topic or asset selectors. A focused watch is easier to evaluate than a broad list of everything.
3. Save it and inspect the initial baseline. Existing evidence establishes the starting state; it is not all automatically a new event.
4. Open What changed. Review source-content changes, dates and coverage before acknowledging the items you have read.
5. If useful, enable selective alerts with an appropriate priority, cooldown and daily budget. Check the matched rule and evidence in each inbox item.

The initial implementation supports up to 20 watches. Typed thresholds bind to the intended field, units and version. Missing or stale evidence does not silently rearm a condition. Cross-watch deduplication and budgets reduce repeated notifications; they do not prove that every relevant development has been detected.

Personal briefings

Choose your timezone and local briefing hour. A personal edition can include a bounded set of qualified highlights, upcoming schedules, forecast outcomes and source gaps. Open the edition before exporting JSON, HTML or PDF. The saved edition is an immutable research record; later source revisions do not silently rewrite its original evidence.

The server scheduler can continue preparing the in-app workflow independently of an open tab. Delivery here is an in-app inbox, not a promise of email or SMS. Browser notification behaviour in other workspaces is separate. Use feedback controls to record whether an alert or briefing was useful, rather than broadening thresholds merely to produce more messages.

An optional research pilot requires an invitation and consent. Participation is separate from plan entitlement; it does not automatically grant additional access or mean that any recruitment target has been filled.

Chapter 07Watchtower Search #

Watchtower Search searches the retained canonical corpus and is available in both Simple and Advanced layouts. It can surface source records, public reporting, social items, data and supported destinations with provenance and freshness context.

Search deliberately

1. Start with an entity, instrument, country or phrase. Add a second discriminating term if the results are broad.
2. Use the available result-type/category filters. A news result, source status row and measured event are different things.
3. Expand the result to read its indexed excerpt and source metadata. The excerpt is what was retained, not guaranteed to be the complete current publisher page.
4. Follow the original source, qualified data detail, country view or market destination.
5. If a Show on map (OpenStreetMap) action is present, read the coordinate qualification. It opens an external marked map for a supported source-owned point; it does not establish a new main-globe observation.

Search is a retained index with a scheduled refresh. A record newly available in a live dataset may appear in Search later. Source-specific filters, retention limits and entitlements still apply. Private My Watchtower items, account records and private investigation contents are not part of the shared public corpus. Live search is withheld during replay rather than being mixed into a historical view.

Use Data Q&A for supported structured numeric questions; use Search when you need to locate records or reporting. Use a specialist workspace when the task requires complete filtered tables, detailed histories or source-native pagination.

Chapter 08News, Social and Narratives #

The News area contains Breaking, Economics, Technology, Health / Bio, Natural Disasters, Search and Narratives. The separate Social workspace is in the main Advanced navigation. Reader topic controls remain available in expanded/fullscreen views.

Read across sources

Choose a topic, date window and geography, then refine by source type, platform or editorial context where offered. Open the original story or post. Compare publication times and publisher families before interpreting repeated coverage as independent support. Reading-list and saved-filter tools preserve the source metadata needed to revisit the item.

Source type, editorial/political leaning, platform and reliability are not interchangeable. A source can be unclassified without being centrist. Wire status does not guarantee neutrality. Labels may be based on publisher self-description or attributed editorial review and do not assign beliefs to every author or reader.

The expanded source directory includes alternative and libertarian publications and selected public Bluesky, YouTube, Mastodon, Lemmy, PeerTube, Rumble-reference, Odysee and podcast sources. These are bounded reviewed feeds, not complete social-platform firehoses. A publisher-linked video reference may retain the referring article's clock rather than a verified video-upload timestamp. A directory-only channel is a discoverable reference, not a connected feed.

Use Narratives to explore themes and related reporting. A cluster describes a grouping, not proof of coordination or truth. When exporting or writing a report, retain publisher URLs, publication dates and relevant source labels. Detailed distinctions are in media methods.

Chapter 09Linked Events and Incident Graph #

Linked Events: follow source records and revisions

Open Linked Events for the feed, filters, timeline, map, associations, sources and recorded changes. Search and narrow by source/type/time before opening a record. Its detail distinguishes occurrence, publication, issuance, source-snapshot and collection clocks where available.

Track a record, add a note or save a view when you want to return to it. The tracker can preserve changes observed since collection began; it does not reconstruct every earlier revision at the publisher. Tracking here is browser-local and does not automatically arm an alert or synchronise to another browser. Export/import or use My Watchtower for its separate cloud workflow.

Associations use shared entities, headline/document evidence and time windows. They are leads for investigation, not causal connections or independent verification. A publisher's country and a place mentioned in text do not automatically establish event coordinates. Relevant unlocated records remain useful in the list.

Incident Graph and Entity Explorer

Incident Graph helps inspect relationships among events, places and related evidence. Select a node or map item, read the relationship description and open the supporting records. Expand the map or description controls when the graph is dense. Ask whether an edge represents a shared reference, inferred association, model relation or observed fact.

Entity Explorer is useful when the subject is an organisation, person, asset or other named entity rather than a single event. Resolve ambiguous names using identifiers and source context. Similar labels can describe different entities; repeated mentions are not evidence of legal ownership or operational control.

A good investigation preserves three things separately: the original record, your relationship hypothesis and the evidence that would confirm or contradict that hypothesis.

Chapter 10Economic Calendar #

The Calendar combines official and provider schedules, reference periods, prior values and market expectations where supplied. It is a planning tool for upcoming information, not proof that a release has occurred.

1. Select the date range and the relevant country, category or source.
2. Read the event's timezone/time basis. A date-only schedule should remain date-only.
3. Distinguish scheduled time, reference period, previous value, consensus and actual release. They describe different stages.
4. Open the official source or event detail for publication and revision context.
5. Use the table or fullscreen view for a longer horizon. Its light-theme headers and day groupings are now explicitly styled for readability.

Compare like-for-like units and periods before interpreting a surprise. A percentage-point difference between actual and consensus is not necessarily a relative percentage change. A passed scheduled time with no actual value is an unresolved data state, not an observed zero.

Calendar evidence also appears as catalysts in Forecast Lab, Geopolitical Outlook and reports. Those appearances do not turn a schedule into a forecast or independent corroborating event. See Calendar methods.

Chapter 11Market Overview, Markets and Global Markets #

The left Market Overview, lower Markets panel and full-width Global Markets now share consistent identities, source dates, units, flags and evidence controls. Use the compact panels for scanning and Global Markets for a wider table or symbol lookup.

Left Market Overview tabs

Lower Markets and the full-width atlas

Commodities shows native futures quotations; Indices shows the registered benchmark universe; Forex offers market quotes and ECB references; Crypto shows the source-defined asset universe; Global Stocks shows the filtered worldwide equity-mover scan. Use the available regional, text, rising/falling, data-gap and period controls. A rising/falling filter requires eligible non-missing change data; an undated source is not automatically a current mover.

Global Markets adds an overview and dedicated Indices, FX, Commodities, Equities, Crypto, Yields and Sources views. Symbol lookup searches supported public references. Confirm instrument type and exchange: the same ticker can identify different listings. Open the instrument for its existing chart controls, or use the information button for normalized evidence and original fields. CSV reflects the selected view; inspect the export's units, source and observation columns.

Worldwide coverage, flags and identities

The benchmark registry contains 56 indices across 44 countries/areas. The market release verified 55 available values, including qualified SET, PSEi, Tadawul All Share and DFM General supplements. Chile S&P IPSA was a visible remaining gap. Availability can change; the registry count is not a guarantee that every record is presently populated.

Index flags identify the benchmark area. Stock flags now prefer source-reported company country, matched to the exact exchange-qualified instrument. That reference is not independently verified legal domicile. If company country is unavailable, a known venue flag carries an explicit Listing tag; if both are unknown, a neutral marker appears. Country is not guessed from quote currency or a bare ticker. Inspect the flag's source/basis and retrieval date; retrieval is not an invented corporate-validity date.

This identity treatment also appears in quote searches, company views, stock-chart headers and portfolio stock rows. Equity display-region filters use the preferred country; the company screener's existing country column is explicitly Listing country. CSV adds company/listing country and flag basis. FX displays the base and quote currency areas; EUR uses a union reference. Commodity mini-icons and crypto glyphs supplement names and tickers. Unknown tokens retain a readable badge. Company-country flag update.

Corrections include the actual S&P/TSX Composite, SSE 50, US CBOE VIX identities and the current CME lumber reference LBR=F. Duplicate aliases for one benchmark are consolidated. An unverified name-only fallback cannot become a newly verified symbol.

Currency-reference example

Open Forex → ECB references, choose a base such as CAD, then inspect a row. The source provides currencies per EUR. Watchtower calculates a cross as quote-per-EUR ÷ base-per-EUR on the same fixing date. The reviewed source supplied 30 currencies including EUR and a 90-day reference file. The prior common fixing supplies the comparison, not an invented intraday quote.

For a USD/foreign pair with a 10% rise, the reciprocal foreign/USD return is approximately −9.09%, not −10%: (1 / (1 + 0.10) − 1) × 100. Pair direction therefore matters in FX strength.

ECB information is available free of charge from the ECB. Cross-rates and changes are Watchtower calculations. Fixings are dated reference information, not executable transaction quotes. See market sources and calculations.

Chapter 12Fear & Greed, Market Tone and bond-move units #

Two sentiment views with different meanings

Fear & Greed restores seven cross-asset scopes—US, Europe, Asia-Pacific, crypto, commodities, forex and bonds—plus a composite, seven regional views and the 44 covered country contexts. The release introduced 59 cards. Search for a scope, open its component detail and check the eligible input denominator and source dates.

These are Watchtower heuristics, not republished CNN, Alternative.me, CoinMarketCap or Tickertape readings. The publisher directory links those services; a directory link does not mean a live numerical integration. The composite uses available core cross-asset scopes rather than adding their regional/country subdivisions again.

Market Tone remains a separate equal-weight regional price-change screen. Its declared transformation is clamp(50 + 12 × mean qualified one-day index percentage change, 0, 100). Missing and stale inputs are excluded with the denominator shown. Neither view is a calibrated event probability or a promise of predictive performance.

Read bond moves in three units

The updated yield rows and details expose basis points, percentage-point change and relative yield change when the source supports those quantities.

Worked example: a yield moves from 4.00% to 4.10%.

• Percentage-point move: +0.100 pp.
• Basis-point move: +10 bps, because one percentage point is 100 bps.
• Relative yield change: +2.50%, because 0.10 ÷ 4.00 × 100 = 2.50.

That last figure is a change in the yield, not the bond's price return. Relative change is withheld when its denominator is unsuitable or the source unit is unknown. Check CSV and evidence details for the same distinction. Release explanation.

Chapter 13Global Bonds, Financial Stress, Deposits and Bank Run Monitor #

Global Bonds

Use Global Bonds for sovereign curves, credit, municipal references, auctions, central-bank context and fixed-income source gaps. Choose the relevant family and jurisdiction, then inspect the exact instrument, maturity, source and date. A yield curve combines tenors; a credit spread measures a difference; an ETF quote is a fund price. They are not interchangeable proxies without an explicit interpretation.

When comparing two curves, match observation dates and comparable tenors. Where official curve changes are supplied, use the stated basis-point/percentage-point treatment. Source-required or unavailable fields identify a data gap, not a normal market. Map references for central-bank headquarters do not show intervention activity.

Financial Stress and Deposits

Financial Stress is the broader finance-research entry point. Use its lenses and drill-downs for funding/liquidity, credit, banking, macro plumbing and related evidence. Open the explanation of a composite before treating a higher value as meaningful. A proxy-based stress index is not an estimated probability of a bank failing.

Deposits focuses attention on bank deposits, reserves, reverse-repo/funding relationships and related source series. Read the reporting lag, institution/system scope, units and denominator. A change in an aggregate series does not identify which individual customers moved money or why.

Bank Run Monitor

Use Bank Run Monitor for the more focused question of banking confidence, deposit pressure and payment-rail evidence. Select a country, institution or evidence family where available. Separate measured financial data from official actions, reported claims and monitoring context. A headline about a bank, a service outage and a reported balance-sheet change have different evidential weights.

For a useful review, write down the original dated observation, whether it is institution-specific or system-wide, any counter-evidence, and what is missing. Use related Financial Stress and country views for context without counting the same source twice. Bank-run evidence methods.

Chapter 14Global Debt and source-anchored debt clocks #

Global Debt covers household, consumer, card, mortgage, business, private, government, total nonfinancial, servicing, arrears, rates and related statistical families. It also retains the earlier Real Estate analysis in its own section. Use the atlas, country dossier, comparison and exact-series histories to understand the selected measure.

Compare a debt measure correctly

1. Choose the sector and measure before choosing the country ranking.
2. Read the reporting period and units: currency amount, percent of GDP, service ratio or borrowing rate.
3. Preserve valuation and lender scope. Market-value and nominal government debt are separate measures.
4. Compare like periods and definitions; do not add household/corporate totals to a private total that already contains them.
5. Export the selected series and source definition when using it outside Watchtower.

Reported, Run, Pause and Resume

The clock shelf starts in Reported mode. Run extends the latest reported stock using the declared four-quarter change and elapsed time. Pause freezes the model instant; Resume catches up. Reported returns to the source stock. A refresh or button click does not create a new observed balance.

The initial clock registry contained 248 models: 242 country measures and six covered-economy subtotals. Those totals cover the source's participating economies, not necessarily all world debt. Clock rates can include exchange-rate effects, valuation changes and revisions; they are not measured second-by-second borrowing or interest.

Running values are gated by comparable consecutive quarters, source age and projection bounds. Missing country measures remain missing. Pins can be displayed in Signal Board; playback is session state, while pins are owner-scoped preferences. For the reasoning behind the animation, see source-anchored debt clocks.

Chapter 15Global Wealth and Affordability #

Global Wealth / Affordability brings together observed price/income/wealth statistics, alternative measures, investment-category references and a real-value lab. Its broader research views cover official inflation, wealth and inequality, national capital, household portfolios, income/wages, housing/rents, food, living standards, debt/taxes, rates/money and resources.

Start with the country and indicator, not an assumed overall wealth ranking. Open an exact series to check its population, sector, frequency and unit. National capital is not household wealth; average wages are not every worker's pay; a rebased housing ratio is not automatically years of income.

Official CPI, annualised one-month inflation, twelve-month inflation and alternative measures have different windows. A Big Mac PPP comparison is a purchasing-power proxy, not an inflation rate. FAO food-price indices describe an international basket rather than a household's local shopping bill. PPP dollars and market-exchange-rate USD should remain separate.

The investment-category directory explains roles and limitations of equities, bonds, cash, housing, REITs, metals, commodities, land, infrastructure, private markets, pensions, human capital, digital assets and collectibles. It does not assign guaranteed returns. Reference-only commercial measures do not imply a subscription or numerical data entitlement.

In the real-value lab, choose explicit nominal-return, fee-drag and inflation assumptions. The real growth factor is (1 + (nominal − fee drag)/100) / (1 + inflation/100). A copied historical CPI value becomes a labelled assumption, not a future inflation forecast. Save the assumptions with the result so another reader can reproduce the scenario.

Chapter 16Crypto Intel and digital-asset research #

The lower Crypto market tab is a quote view. Crypto Intel is the deeper research workspace, with Overview, Markets, Stablecoins, DeFi/venues, Chains, NFTs, Derivatives, Network, Security, Global policy, News and Sources. Existing model/watchlist/analyst functions sit alongside this source-based research.

Search and filter the chosen family, open an asset or protocol, inspect a history and use comparison/export controls where offered. Complete filtered exports can span multiple pages; preserve a coherent source snapshot. Provider rate limits and partial directories remain visible.

Important distinctions

• Stablecoins: token supply and USD-valued circulation differ. A non-USD peg requires its proper FX basis; yield/share/fund tokens are not automatically USD1 pegs.
• DeFi and venues: TVL, trading volume, fees and revenue are different quantities. Overlapping protocol and chain totals should not be summed blindly. CEX asset references are not proof of solvency.
• Yield pools: APY is source-reported context, not a guaranteed forward return.
• NFTs: a listed floor is not a bid; floor × supply is not readily liquid market value. Addresses are not people. Collection directories can be partial.
• Derivatives: preserve contract type, expiry and native open-interest units. An options notional and a contract count are not the same measure.
• Security and policy: distinguish a historical exploit report, an official regulatory reference and an issuer assertion. A directory entry is not a legal approval or reserve audit.

Use the Source view to understand coverage before comparing chains or countries. A country filter in policy references is not a physical-risk map of a blockchain.

Chapter 17Monetary Predictions, Gold and Commodity Wars #

Monetary Predictions contains several related but distinct research areas. Use the internal navigation to move between monetary/hyperinflation context, Gold and Commodity Wars 2030. Read the selected model's input definitions, source periods and evidence gaps before interpreting a ranking.

Monetary and gold research

Monetary screens combine the source evidence and declared methodology available in the workspace. A stress ranking or authored estimate is not automatically a calibrated probability of hyperinflation. Compare the underlying inflation, money, exchange-rate and fiscal observations rather than only the headline score.

Gold research links bullion/currency context, official reserve references and mine-production evidence where supplied. Separate a reserve stock from annual mine production, a futures quote from physical-dealer pricing, and tonnes from currency-valued holdings. Country references do not imply that all reserves have been independently audited by Watchtower.

Commodity Wars 2030

This is a source-based country resource/planning atlas. Choose a metric, reference year and estimate policy, then compare countries on the same measure, basis, unit and year. The reviewed catalogue initially contained 309 separate measures and a critical-material applications directory. “2030” is planning branding, not a forecast that a commodity conflict will occur in that year.

Production, reserves, resource endowments and processing capacity are different measures. Freshwater resources and adverse water stress point in different directions. Missing first-stage quantities remain unscored. A same-basis 0–100 ranking is relative to its eligible set; it is not a probability or a universal resource-security score.

Use Commodity Atlas to inspect physical references and Companies to investigate issuers, while retaining the distinction between national production statistics, mapped sites and company disclosures.

Chapter 18Prediction Markets #

Prediction Markets → Market Radar provides a risk radar, screener, repricing view, closing calendar, event ladders, themes, browser-local watchlist, comparisons and source methods. The integrated catalogue is a reviewed subset of Polymarket and Kalshi, not each venue's entire market universe.

Open the exact contract before interpreting a displayed price. Read the outcome wording, Yes polarity, resolution criteria, close time and quote time. Two similar headlines can have different settlement conditions. Nested thresholds or contracts about the same event overlap and should not be added into a total probability.

Use books/depth and histories where available to understand the quotation's context. Missing trades or a thin book should remain missing/thin, not be replaced with an assumed price. USD volume on one venue and contract counts on another are distinct measures. Related news or a Watchtower model can provide context, but does not settle the contract.

A market price is a market observation about a particular payoff. It is not Watchtower's calibrated forecast or confirmation that the event happened. Watchlist saves here are browser-local. See Prediction Markets methods.

Chapter 19Companies and issuer research #

Open Companies when you need to move from a commodity theme or ticker to a specific issuer. Search the reference universe, identify the correct listing/entity and inspect the source-backed dossier. Use the available source, category, geography and history controls rather than assuming every result is an active listed company.

The mining/extraction universe combines selected SEC extraction/mining registrants, Wikidata entities and labelled curated references. Broad industry categories can include energy and services. Company names alone do not establish commodity exposure; a “junior” label or low market capitalisation does not establish operating stage. Listing-venue and business-address country are not mine locations.

For a selected supported SEC issuer, primary facts retain the entity identifier, taxonomy concept, value, unit, reporting start/end, filing date, form and accession source. Compare compatible periods and currencies. An amendment can supersede a fact for the same period. The panel does not manufacture trailing-twelve-month results, EV multiples or cash runway from unmatched observations.

Production, reserves/resources, AISC, guidance, mine ownership and project capex require current issuer disclosures or technical reports. Use their source links for those questions. A reference company population is useful discovery, not an exhaustive current worldwide mining census. Optional connected-provider data remains subject to your account and the provider's licence. Companies methods.

Chapter 20Countries and Global Politics #

Countries: build a dossier

Choose a country in Countries, then work through overview, statistics/trends, encyclopedia material, source references and Watchtower handoffs. Compare up to four countries where the comparison tool is offered. Export exact series with their years and units, rather than taking a screenshot of a headline without its period.

The reference catalogue covers 250 country/territory identities, with uneven evidence by indicator. World Bank statistics and archived profiles can have substantial reporting lags. Wikipedia prose is an on-demand, revision-dated reference. The CIA Factbook was retired in February 2026; archived fields keep their original years and should not be described as a newly updated live Factbook.

Use View in Watchtower links to move to relevant modules. The destination may apply a country filter, focus a map, run a search or open a global view. Read that handoff's scope; it does not guarantee that every destination has country-specific observations.

Global Politics: compare governance evidence over time

Use Global Politics for its country reference views, indicators, history/trends and comparisons. Choose a measure, source year and coverage set before comparing rankings. Governance, democracy, development and economic observations have different definitions and scales. A change between vintages can reflect source revision as well as real-world change.

Separate the published indicator from a Watchtower-derived comparison and from your interpretation. A country with no qualified value is not automatically the best or worst performer. Use Countries for the broader research dossier and Geopolitical Risk for current source-linked event/model context.

Chapter 21Geopolitical Risk and Outcome Tree #

Geopolitical Risk / Outlook

The workspace includes Situation/Outlook, the Country table, forecast exploration, evidence timelines, market expectations, catalysts, track record and sources, with classic diagnostics retained separately. Select a country or region and a horizon, then open the evidence behind the chosen view. Fullscreen map-and-evidence mode keeps the map with collapsible filter and data rails.

Country tables contain source-year fiscal, market, governance and model fields. Check their individual periods and qualifications. A global scenario displayed beside a country does not become that country's probability. Reporting attention describes publication activity in its stated windows; it is not an incidence rate or coverage-adjusted measure of danger.

Compare up to four countries where offered, but keep a consistent indicator definition and date basis. Prediction-market contracts and Calendar catalysts remain separately defined evidence. Optional historical conflict layers are dated context, not a current census. Geopolitical Outlook methods.

Outcome Tree / Global Risk Studio

Outcome Tree makes the authored cross-domain scenario engine inspectable. Its eight non-exclusive scenario families span banking/liquidity, cross-asset dislocation, strategic supply corridors, grid/digital infrastructure, geopolitical escalation, health/biosecurity, mobility/official controls and food/energy pressure. Supported global horizons include 24 hours, seven days and thirty days.

1. Select a scenario and horizon. Read the precise criterion and forecast window.
2. Inspect retained contributors, excluded/context-only inputs and dependency groups.
3. Open the original source records and their clocks. Shared inputs can make models dependent.
4. Use a hypothetical experiment to remove groups, attenuate evidence or change an assumed prior. Observe sensitivity without treating it as a real intervention.
5. Inspect the track record and resolution state before interpreting a numerical estimate as demonstrated predictive skill.

These scenarios are not mutually exclusive, so their values do not sum to 100%. Published conditional pathways are shown only where the source model supports them. Compound-risk views use dependence-agnostic mathematical bounds rather than assuming independence. A sensitivity range is not automatically a confidence interval.

Saved experiment recipes are account-local browser preferences. Reapplying a recipe to a newer capture can produce a different experiment; read the changed-capture message. An active experiment may stay on its accepted capture until you explicitly refresh. Outcome Studio methods.

Chapter 22Bayes Threat, Ancap Welcomeness and country taxes #

Bayes Threat · Ancap Welcomeness now opens with country scores and independently refreshed evidence. Its country search, regional selection, comparisons, dossiers, pillar audit and JSON export provide several levels of detail.

Understand the score before ranking countries

The declared model uses twelve value pillars and a 60/40 liberty/practical balance. It is a value-based comparative model, not a probability that an event will happen or a personal guarantee about relocation. All 250 country/territory references can receive numerical estimates, but ranking eligibility depends on data-backed coverage. A neutral placeholder for missing evidence is not an observed average country outcome.

Select a country, open its pillar and source/weight breakdown, and inspect the reporting/review periods. Missing-input ranges are sensitivity bounds, not statistical confidence intervals. Related WGI and V-Dem evidence should not be treated as independent confirmation merely because it appears in separate rows. Historical comparisons must retain their model version.

Taxes & fees

The Taxes & fees view brings together source-reviewed professional tax summaries, World Bank fiscal/fee histories and selected official registry fees. Choose the country and tax category, then read the jurisdictional scope and conditions. Personal and corporate taxes, inheritance and gifts, and federal versus local charges are distinct.

A headline VAT or income-tax rate is not an effective tax bill for every person. A publisher's review date is not automatically a law's effective date. Conditional schedules should not be reduced to their first number. Historical Doing Business benchmarks remain historical; a listed incorporation fee applies to its stated service/channel and currency.

The declared fiscal screen uses fixed weights and explicit missing-input bounds. Unsupported local, social, sectoral, residency, relief and fee coverage remains a gap. Use the original authority/professional source for a specific legal or tax situation. Welcomeness methods · Tax and fee methods.

Chapter 23Risk Overview and Systemic Risk #

Risk Overview is a cross-domain organising view for conditions, evidence and related context. Use its selected lens, map/rows and explanations to decide which specialist workspace to open. Distinguish an assessed model state from the number of source records available to support it.

Systemic Risk adds a deeper, typed research workflow with Current, Outlook, Impact & dependencies, Evidence, Track record and Methods. The same systemic research is available inside Forecast Lab. The current framework describes fourteen domains, including recession, while preserving different units and assessment eligibility.

A disciplined systemic-risk review

1. Begin with Current. Check which domains have assessable numerical inputs and which remain incomplete.
2. Inspect Evidence for the original observation/publication/source clocks and the reason an input is eligible or excluded.
3. Open Outlook and choose the exact financial or regional target, model and horizon. Read the target definition before comparing probabilities.
4. In Impact & dependencies, distinguish an observed exposure share from an assumed relationship. A graph edge alone does not justify a numerical loss propagation.
5. Open Track record. Separate published-vintage historical research from genuinely issued prospective forecasts and compatible resolved outcomes.

The financial expansion defines 58 terminal-loss, from-base close-path-loss and joint-decline targets across eleven US-listed fund references at one, five and twenty NYSE sessions. Those are specified fund-price outcomes; changing a geographic-context selector does not turn them into local-country crisis probabilities. Historical reconstruction has used a longer common source record beginning in 2007, with versioned methods and baseline comparisons.

Regional warning research concerns sampled severe/extreme Actual/Public CAP-message onset in its specified UTC window. It is not a direct physical-hazard or outage probability. A negative label requires sufficiently complete healthy sampling; a gap in collection cannot establish that nothing happened. Specialist targets may remain ineligible until sufficient suitable history exists.

Current model indices are partial proxies. Research estimates do not become proven live forecasts merely by appearing in a dashboard. Promotion requires meaningful prospective evidence, calibration, baseline improvement and utility under the declared criteria. Systemic Risk methods.

Chapter 24Forecast Lab and Energy outlook #

Markets, scenarios and track record

Forecast Lab combines statistical market research, source-linked cross-domain scenarios, weather outlooks, economic catalysts, conditional comparisons, personal browser watches and immutable forecast/evidence records. It also includes the newer Systemic risk and Energy views.

For the original Markets pilot, the nine reference funds are SPY, QQQ, EFA, EEM, TLT, HYG, GLD, USO and UUP. Targets are defined over one, five and twenty NYSE sessions. A displayed directional estimate refers to the stated fund-price return, not every constituent, a country, a spot commodity, a yield or a total-return investment strategy. Distributions, execution costs and slippage are not silently included.

Select an instrument and horizon. Read its base and target session, empirical range, component/analogue evidence and baseline. Then inspect historical and prospective performance separately. Brier score, log loss, calibration and interval coverage answer different questions; a nominal 80% interval still needs observed coverage evaluation.

Live issuance uses actual issue times and freezes its definition, source cutoff and inputs. A missing target session remains awaiting evidence. A source revision can withhold scoring instead of rewriting the original forecast. Previously displayed research is not automatically a backdated issuance.

What-if comparisons and local watches

The What-if Lab selects non-overlapping historical windows meeting a specified reference-fund return condition and examines another fund in the same windows. It needs enough matching cases. This is conditional historical co-movement, not a causal shock-response estimate or a calibrated future conditional forecast.

Forecast Lab's original watches and saved baselines are account-scoped in the browser. Their visible-workspace estimate-change checks do not imply closed-browser delivery. Use My Watchtower's separate server workflow when that is the intended monitoring channel. Forecast Lab methods.

Energy

The Energy view includes price overview, futures curves, production/impact context, a declared supply scenario, sources/coverage, historical observations, research ranges and exports. It brings together EIA, World Bank, EU consumer-price and selected electricity references with native currencies, units, taxes and periods.

Select the series and market before comparing numbers. USD per barrel, USD per MMBtu, EUR per litre, monthly commodity benchmarks and hourly/daily electricity references describe different things. Published futures quotes are delivery-contract references, not expected future spot prices or official settlements unless explicitly identified as such.

The research projection uses a fixed declared historical method and persistence comparison. It is not established live forecasting skill. Current news/weather is context and is not inserted into historical training as though known then. The supply scenario uses explicit local-loss and elasticity assumptions; a missing comparable world total prevents numerical propagation. Production-country shares retain their source years. Energy methods.

Chapter 25Infrastructure and Pacific/subsea networks #

Infrastructure is the physical/network reference workspace: use the available asset and network layers, search, map inspection and source controls to investigate power, communications, data centres, exchanges, pipelines and related systems. Distinguish a mapped reference network from a measured service outage or current flow.

Pacific cable connections

The recent update makes the map wrap across the date line and adds a Pacific cables view. The selected FASTER, Unity, JUPITER, PLCN and PC1 connections are drawn as purple dashed schematic links between approximate, system-specific localities. They describe same-system connectivity, not surveyed seabed paths, precise landing locations or inferred cross-connects between neighbouring systems.

Open a line or locality, read the system/source description and use the original operator or specialist reference. Omitted branches and source history remain explicit. A reference to a cancelled route does not make it part of the current schematic; the cancelled PLCN Hong Kong branch is excluded.

Underwater power and pipelines

Use the Undersea networks controls for power/pipeline layers, search, lifecycle/history and Pacific/world framing. The parent layer matters: power follows Power or Global Grid, pipelines follow Pipelines, and schematic communications follow Fibre/Cables. If an overlay is unexpectedly absent, check the parent toggle first.

The October OSM reference capture contained 3,260 power ways and 15,332 pipeline ways, including inland-water crossings. These are mapped ways, not unique complete systems or an exhaustive operating inventory. Construction, proposed, disused and abandoned tags affect the default view; unverified lifecycle is not assumed operational. Shared-node components describe mapping topology, not proven live flow.

Native infrastructure, atlas overlays and these supplemental references have separate counts and scopes. Preserve OSM/ODbL or the stated schematic attribution in exports. Subsea methods · Fibre methods.

Chapter 26Commodity Atlas, Commodities Data and gas facilities #

Commodity Atlas is the map-led resource workspace; Commodities Data is its data-oriented companion. Use commodity/category, geography, source and available lifecycle/history filters before selecting a site, network or country reference. Open details for the native identifier, original source, observation/reference date and geometry meaning.

Commodity prices, national production totals, reserves/resources, processing facilities, ports and transport links are different layers. A mapped mine reference does not prove current production, company ownership share or available capacity. A country production series should not be allocated to a site without an explicit source.

New industrial-gas references

The gas-facility addition contains eleven curated facility/complex records and thirteen gas identities at release. It supports process, gas and text filtering and includes a Neon shortcut to the source-backed POSCO AIR SOLUTION Gwangyang reference.

Explore air/rare gases, helium, natural gas/LPG, electrolysis, reforming, gasification, CO₂ and biogas processes as represented. Specialty-synthesis gaps are explicit. Source-reported planned output is separate from operating output; an undated source keeps its undated status.

Markers are authored approximate locality references, not verified facility coordinates. Read the dashed/reference labels and source cards before using the location. The inventory is partial. Follow a reference to Companies for issuer research and to Monetary Predictions → Commodity Wars for same-basis country resource comparisons. Gas-facility release.

Chapter 27War Room: monitoring, investigations and terrain tools #

War Room brings a quiet map, source-linked monitoring, saved investigations, imagery and explicit terrain scenarios into one workspace. Its main modes are Monitor, Investigate and Tools.

Monitor a place and time window

Choose a place, coordinates or an area of interest, set a time range and select relevant evidence sources/types. Use the synchronized map and list; on compact screens, switch among Map, Events and Details. Original source clocks and unlocated evidence remain important. A new-to-collection count is not automatically a newly occurred event.

Open a record to inspect the source and location basis. Publisher-country metadata is not event geolocation. An area's bounding box is a filter, not an impact footprint. Use the source's own warning or observation geometry when available.

Build an investigation

1. Name the question and record a tentative assessment.
2. Pin source snapshots, retaining stable citations.
3. Mark evidence as supporting, contradictory, contextual or unresolved.
4. Compare a bounded set of records, write notes and list gaps.
5. Save the case in the supported browser-local account space; export JSON for backup or cited Markdown/printable HTML for a report.

The current case workflow supports browser-local saved cases, not automatic cloud synchronization. Pinned snapshots preserve what you selected; they are not guarantees that the source remains unchanged. A manual “reviewed” status is your annotation, not an automatic factual-confirmation engine.

Imagery and terrain

Esri imagery can combine acquisitions of different dates. Sentinel-2 scene previews carry acquisition times and scene-level cloud context. NASA GIBS provides dated daily mosaics with native resolution/coverage limits. Zooming in does not create higher-resolution source information. See workbench imagery methods.

Under Tools, area analysis can show slope, aspect, contours, relative low ground, drainage/wetness proxies and terrain visibility. Choose the site/study width, run the calculation and read grid spacing and valid-sample coverage. A coarse terrain screen omits buildings, vegetation and sub-grid features.

The communications/antenna tools use user-selected endpoints, heights, frequency and clearance assumptions. Their output screens terrain/Fresnel clearance; it is not a complete radio link budget. Nearby alternatives are the best among sampled candidates, not a global optimum or a finding about property access.

The dam scenario uses chosen toe/down-valley points and assumed stage rises to explore terrain-connected cells. Its coloured variants are assumptions, not occurrence probabilities, wave arrival times or safe evacuation routes. Boundary contact means the result may be clipped. Export the assumptions with the result and use the terrain/planning methods for exact definitions.

Chapter 28Conflict / Military #

Conflict / Military aggregates source-qualified conflict and military-related public evidence. Use geography, time, source and layer filters, then select a record and read its geometry, event class and original report. Country outlines, event points, reference facilities and reported activity should remain distinct.

The unified map is useful for moving from a regional overview to the underlying source, rather than drawing conclusions from the density of markers alone. Multiple reports can concern one development; uneven reporting can leave significant gaps. A military-related headline, an official statement and an independently observed event have different meanings.

Use Geopolitical Risk for global model/context views, War Room for a place-based investigation, Aviation for its dedicated public aircraft observations, and Satellite & Sensor Data for orbital/reference context. A source military label on an aircraft is not independently verified ownership, mission, occupants or intent. Sharing a location or date across layers is a research lead, not by itself a causal or operational conclusion.

Chapter 29Air / Aviation #

The dedicated Air / Aviation workspace supports 2D/3D views, unified search, filters, aircraft detail, follow, trails/charts, airport lookup, replay, comparisons, context, licensed reference images, private investigations and supported cloud views/watches/inbox functions.

Follow a public observation

1. Search by a supported callsign, registration, identifier or airport reference. Check the identity fields rather than relying only on a familiar label.
2. Select the aircraft. Compare position time, last contact time and collection time.
3. Inspect altitude/speed/track units and the available observed trail. A directional marker follows supplied track; missing direction is not invented.
4. Use follow or fullscreen when useful. Switch to retained-observation replay for history actually captured by Watchtower.
5. Export the supported public observation or trail with its source attribution and gaps.

Public ADSB.lol is the default source. Receiver/query/metadata coverage is uneven; counts include source targets and are not all worldwide flights or verified flight legs. Retained history grows from actual collection and can shorten under caps. Gaps and implausible jumps split trails rather than being joined into a fabricated flight path.

Airport, weather and airspace context

Context can include OurAirports references, AWC METAR/TAF/SIGMET/G-AIRMET, FAA status/TFR/airspace and retained weather warnings. FAA coverage is not a complete global NOTAM service. A weather polygon overlapping a horizontal aircraft position does not prove flight impact. Read original notices and validity windows.

Personal and licensed additions

Use the workspace's supported cloud views, watches, server inbox and account-private investigations for its newer operations workflow. Browser notifications require the relevant open browser context; server inbox evaluation is separate. Local watch preferences and cloud records are not interchangeable.

API Connections & Add-ons → Aviation Data Add-ons can use your eligible FlightAware AeroAPI, ADSBX Enterprise or aviationstack credentials. These are manual, owner-private lookups with separate provider terms/charges. They do not turn paid results into shared public trails, watches or cases. Aviation methods · Investigation release.

Chapter 30Marine and Mobility Intel #

Marine

Use Marine to investigate the available vessel, port, shipping and chokepoint references. Inspect the source and geographic coverage before interpreting a ship count. Public AIS samples may be regional, delayed or incomplete; a last-known position is not a current confirmed location. Embedded third-party views retain the provider's own coverage and access.

Compare a port or route observation with its source time, baseline and relevant event evidence. A position or traffic count does not directly establish cargo, ownership, delay, capacity utilisation or cause. Use Commodity Atlas and Infrastructure for physical context without treating nearby reference assets as proven affected assets.

Mobility Intel

Mobility connects ports/chokepoints, aviation weather/status and selected rail, border or corridor evidence. Its direct-evidence views retain native records, scope and formulas. PortWatch activity, METAR weather, FAA restrictions and sampled Finnish rail observations measure different things.

A weather report is not an airport-congestion reading. A scoped FAA restriction does not imply an airport-wide closure. Predicted rail timing is distinct from observed lateness. A same-weekday activity baseline needs adequate comparable samples; insufficient or stale data should not become an infinite anomaly score or a quiet condition.

For an investigation, choose one node/corridor, inspect its original operational series, then add weather, infrastructure and reporting as separately labelled context. Distinguish reference corridors from measured operations and preserve that distinction in exports.

Chapter 31Weather Alerts and Natural Hazards #

Weather Alerts

The global warning view offers country, severity, current/upcoming and multilingual text filters; paginated lists; original information/instructions; source details; map geometry; JSON export; and a country/territory coverage directory.

Read the issuing authority, issue/effective/expiry times, severity and message status. CAP updates and cancellations follow explicit source references. Multiple messages or language blocks are not necessarily separate storms. Japan's JMA municipality/hazard states retain their own levels and are not simply interchangeable with CAP severity categories.

A coverage directory listing does not guarantee complete current delivery for that country. No reported warning, partial, stale, not integrated and unavailable have different meanings. Warnings without qualified geometry remain in the list; Watchtower does not fabricate a country-centre impact polygon. Global warning release.

Natural Hazards

Use Natural Hazards for earthquakes, severe-weather products, fires, tropical systems, volcanic evidence and related observations/reference layers. Its recent expansion reuses still-valid global warnings, adds source event inspection and improves native-unit temperature, rain and snowfall cards.

The volcanic controls include source bulletins and published VAAC observed/+6/+12/+18-hour airborne-ash areas, with source flight levels. Airborne ash is not ground ashfall, a pyroclastic-flow footprint or an evacuation zone. An older open EONET listing does not establish a continuing eruption. No synthetic danger cone should be inferred from a volcano point.

Optional historical landslide reports, dated fault traces, reference volcanoes and LHASA model imagery have separate meanings and are generally off until selected. COOLR reports can be historical. GEM fault traces are geological references, not earthquake predictions. Provider-latest LHASA imagery may lack an exposed run timestamp.

Snow depth appears only when a source supplies it; rainfall is not converted to snowfall with an assumed ratio. Preserve observation/forecast/maximum labels, units and accumulation windows. The newness highlight uses an eligible source time for a bounded ten-minute window; refreshing does not restart an event's age. Expanded hazard methods.

Chapter 32Space Weather and Satellite & Sensor Data #

Space Weather tabs

The Space Weather area contains Live, Threats, Forecast, Aurora, Asteroids and Sources. Use it to distinguish measured solar-terrestrial conditions, official NOAA products, research models and NASA/JPL orbital references.

Live/Threats can include Kp, solar wind, radiation channels and official R/S/G scale context. Read channel, unit, quality flag and validity time. Kp, local K, Dst and geoelectric fields are different quantities. Missing official values should not be interpreted as G0/R0/S0. A regional geoelectric model is not a measured induced-current or power-failure map.

Aurora uses model fields and their valid times, not a direct sky observation. Inspect a cell, frame date and hemisphere/world view. Cloud, local terrain and light pollution affect what a person sees and are not automatically represented. GloTEC background-only versus observation-supported cells have their own support meaning.

Asteroids separates orbital catalogue references, predicted close approaches, Sentry monitoring, Scout candidates and historical fireballs. MOID is a distance between orbital paths, not a predicted simultaneous encounter. PHA classification is not an impact prediction. Source orbital epochs and TDB approach dates differ from a present-time position. Space Weather methods.

Satellite & Sensor Data

Search the orbital/reference catalogue, filter ownership/type/coverage, select an object and inspect its source elements, epoch and projected track. Positions calculated from orbital elements are estimates with age limits. Missing or incompatible elements do not acquire invented positions.

Ownership references can be uncertain or use classifications that are not simple ISO countries. An organisation attached to a station module is not necessarily the owner of the whole station. Object counts can include reference/uncertain records and are not a count of confirmed operational satellites. Layers and fullscreen controls change presentation while retaining these qualifications.

Chapter 33Outages and Cyber Exposure #

World Disruption Observatory

Outages brings together service incidents and other disruption-related evidence with source/time/location filters, maintenance/recovery/context/stale controls, a map, an unlocated list and exports. Open the source detail before treating a coloured marker as an ongoing failure.

A provider status notice, planned maintenance, a recovery message and a measured regional power fault are different records. An unlocated cloud-service event should not acquire a headquarters “impact point.” Satellite orbit positions are not customer-impact footprints. Old standing-status notices are not newly failed systems. Source record counts are not unique outage counts.

Cyber Exposure

The deeper cyber workspace includes advisories, CISA KEV, EPSS, recent critical CVEs, package advisories, Internet measurements, response directories, ATT&CK, reporting, dated C2 history and sources. Select the view and identifiers, inspect native evidence and export a coherent filtered selection.

CVSS describes technical severity; KEV records known exploitation under its catalogue rules; EPSS is a defined thirty-day exploitation estimate/rank. A due date can apply to a particular regulated audience. None of these alone proves that your organisation runs a vulnerable affected version.

Country lenses may show measurement samples, issuing jurisdictions, directory references or report associations. They are not automatically attack-origin/victim locations or national cyber-safety rankings. Historical hosting references remain historical even if a retained source flag once said “online.” Use the original advisories and affected-version detail for a concrete exposure investigation.

Chapter 34Sanctions / Ownership and CBDC & Digital Money #

Sanctions, trade controls and ownership

Open Sanctions / Ownership for listed targets, entities, individuals, vessels/aircraft, trade/export controls, country/thematic regimes and sources. Search for an identifier and open the exact source entry. Preserve the list/programme, measure, reasons and dates.

List entries can overlap and are not a deduplicated population of legally resolved identities. Similar names do not establish identity. Blocking, sectoral restrictions, banking measures, licensing requirements, verification lists and procurement/import rules are distinct. A source programme's update date is not automatically a designation date.

Country programme maps do not establish a universal embargo or a restriction's severity. Ownership search/reporting remains a separate research tool; the workspace does not automatically resolve ownership chains or apply every jurisdiction's ownership rule. Use the issuing authority's current record for a specific case.

CBDC & Digital Money

Explore country ratings, retail CBDCs, wholesale money, private money, cross-border projects, regulation, stablecoin context, dated evidence, comparison and source gaps. Select the track before comparing countries.

The declared maturity bands run from research through design/prototype, controlled pilot, limited operation and operational product. They are ordinal deployment stages, not percentages of national readiness, adoption, privacy, safety or investment merit. A country rating is based on eligible explicitly mapped project evidence within one track.

A shared project can cover several jurisdictions without becoming several independent programmes. A completed historical pilot is not necessarily current operation. Private stablecoins and tokenised deposits are not CBDCs; supported currency or partner participation does not establish national adoption. Editorial review and source availability expire separately, so a fetched page does not automatically renew a rating.

Chapter 35Insurance Risk & Resilience #

The navigation tab Insurance Stress opens the broader Insurance Risk & Resilience workspace. Its views include country/industry statistics, solvency/capital, premiums/claims, balance sheets, allocation, market development, protection gaps, backstops, catastrophe observations, selected insurer references, histories, comparisons and sources. The older US proxy monitor remains separately labelled.

Choose a financial measure and reporting period before using a map or ranking. Preserve currency multipliers, ratio scale, solo/group scope and provisional/final status. A ratio of 1.0 can mean 100% in the source; EUR million is not EUR. Year-to-date quarters are not independent flows to be summed. Country reporting domicile is not an insurer's entire underwriting footprint.

Source-qualified review flags identify declared thresholds, such as a reported solvency or combined-ratio condition. They are not blended insolvency probabilities. A net combined ratio uses its specified claims/expense and premium basis; investment income and a claims-paid/gross-written-premium calculation are different.

Catastrophe points are source physical-event references. They do not automatically establish insured losses or carrier exposure. EIOPA protection-gap models, historical statistics and dated scheme descriptions retain their vintages. The selected insurer directory is a reference collection, not a complete licensed-register census. Retain source/derivative/free-original-material notices in exported work.

Chapter 36Health / Biosecurity and Crime & Terror #

Health / Biosecurity

Use Health's evidence views for official disease/outbreak reporting, wastewater and animal-health context where supplied, plus its facilities reference view. Select the source, country, date and event type, then read reporting/observation periods and geographical coverage. More reports can reflect more surveillance rather than a directly comparable increase in disease incidence.

The Facilities view supports source, type, focus, containment and lifecycle filters. It combines reviewed institutional references with supported registration evidence. FDA registration is not product approval or containment certification. BSL4, CL4, PC4 and other source schemes retain their own labels. Planned or commissioning facilities do not become verified operational capacity.

Facility markers can be city/locality references rather than building coordinates. Unmapped records remain available. A source reference should not be interpreted as a complete census of laboratories, manufacturing plants or national health capacity. Health methods · Facility methods.

Crime & Terror

This map-led workspace combines global evidence with selected police/agency bulletins and statistical area/cell references. Use the filter and data rails, source/lag details, unlocated records, comparison and export tools. Some workflows open a map-first fullscreen view; close it to restore the original tab.

Police bulletins can be preliminary. A municipality centre is not an incident site. Statistical cells aggregate a specified category/month and reporting sample, while coverage can be capped or lagged. Different jurisdiction definitions and collection windows prevent treating raw counts as comparable personal-risk rates.

Inspect the original classification, period and geometry before combining Crime & Terror with a country dossier or a War Room case. A blank area can be uncovered or suppressed by filters; it does not establish low crime. Separate your assessment from the source's statement.

Chapter 37Cameras, viewing walls and Flock locations #

Public camera explorer

Open Cameras to search a place, radius, country, source or media type. Inspect the source record before opening an image or stream. Official snapshots, community discovery links and directory-only references are different. A registered camera record does not guarantee live playback.

Create a viewing wall using the supported 1/4/9/16-slot layouts, favourites and explicit start/stop controls. Custom public HTTPS images/HLS/clips or YouTube links may be supported, subject to browser, codec, CORS and provider rules. The wall does not create a recording or a media-rebroadcast entitlement.

Walls/favourites and custom URLs are account-scoped browser preferences. Export/import them for a deliberate backup. Hidden/replay/account transitions stop or clear playback as required; opening a new account should not restore another account's wall. Camera methods.

Flock locations

Choose Flock locations for the community-mapped OSM/DeFlock ALPR reference inventory, its map/list/search/radius filters, explicit bearings and source-change history. This is a database of reported locations, not plate records, camera video, verified hardware counts or current operational status.

An added or removed source record is a mapping change; it does not prove that hardware was installed, removed or moved at that instant. No viewing cone is inferred from a point. Tracked IDs are browser-local and do not send notifications. Preserve the source dates and ODbL attribution in exports. Flock release.

Chapter 38Portfolio, personal relevance and exposure monitoring #

Portfolio applies research context to the holdings and exposure information you choose to maintain. Begin by checking the verified account and the saved portfolio state before editing. Identify each instrument and listing, enter quantities/cost assumptions in the appropriate units, and confirm the application's save result. A portfolio view is not a broker order ticket.

Inspect quote freshness and mapping coverage before interpreting valuations or exposures. A fund can hold many underlying assets; a ticker's listing country does not establish its economic exposure. A missing quote or unresolved mapping is a gap, not a zero-value holding. Snapshot valuations, historical research and scenario assumptions should remain separately labelled in any report.

Personal-relevance/consequences tools can connect selected evidence to your supplied context. Treat their results as conditional research: the source event, exposure mapping and assumed consequence are separate steps. A geographic overlap or shared industry label is not measured loss or a guaranteed impact on a holding.

Where the account's exposure-monitor rules are available, inspect the rule's fields, thresholds, evaluation status and delivery channel. These rules have their own server-side contracts. They do not automatically convert every browser rule, watchlist or map pin into unattended monitoring. Export the assessment assumptions with your evidence when sharing a portfolio-related conclusion.

Chapter 39Alerts and Intelligence Rule Studio #

There are several monitoring workflows. Choose by where the condition is evaluated and where you expect to read the result.

Build a multi-input browser rule

1. Open Alerts → Intelligence Rule Studio and name the question the rule is meant to detect.
2. Select inputs from the available loaded public-source field catalogue. Read each field's units and timestamp basis.
3. Combine conditions using ALL (AND), ANY (OR) and optional NOT groups. For example, A AND (B OR C) makes a different claim from (A AND B) OR C.
4. Set the maximum age, matching/crossing/duration behaviour and cooldown deliberately.
5. Test against the current evidence. Inspect true, false and unknown branches before saving and arming.
6. Read a triggered report for the exact thresholds, source values, times and optional scenario note.

Missing, stale, failed or future inputs are unknown. NOT unknown remains unknown. A crossing needs two comparable timestamped values. “Observed true for” means matching sampled checks over the chosen duration; gaps can reset that duration. Unknown data does not silently rearm a condition as if it had become known false.

Imported definitions require review and arming. Browser-local rule/report exports can contain your notes and assumptions. Experimental linear projection from session samples is an extrapolation of observed snapshots, not a calibrated event forecast or invented historical series. Rule methods.

Chapter 40Data Q&A, AI Analysis and Oracle #

Data Q&A

Use Data Q&A for supported structured questions over reviewed public sources. Its Ask, Dataset catalogue, Evidence, Browser snapshots and Methods views help you inspect what was used in an answer.

Ask narrowly: specify a country, measure and period where appropriate. For example, ask for the source-reported household debt measure for a particular country rather than “Is this country safe?” Inspect the answer's citations, source-defined units and reporting periods. Some questions produce a source summary and a link to a deeper workspace rather than a fully reconstructed analysis.

The catalogue's registration, a checked response, a source collection and an original observation are different states. Dataset counts and browser snapshot sizes are coverage diagnostics, not independent observations. The exported answer includes its supported displayed evidence scope; it is not necessarily an entire upstream dataset.

Data Q&A is a deterministic evidence lookup workflow, not a new generative forecast model. Its answer history and draft context have their own session/owner lifecycle. Use the AI handoff explicitly when you want synthesis.

AI Analysis and Oracle

AI Analysis provides entitled question/SITREP-style research over available evidence. Oracle is a separate top-bar entry point. Check the selected tool's current capabilities, source scope and any displayed credit/provider cost before requesting work.

A useful prompt has a question, geography, period, requested output and evidence standard. For example: “Summarise the latest source-dated energy and shipping evidence for this region; separate observations, model context and missing coverage, and cite the source dates.” That is a request format, not a claim about current events.

Review generated text against the cited records. An AI summary can help organise evidence but does not independently verify an event, supply missing source coverage or establish model calibration. Ask for explicit uncertainty and alternative explanations when the data supports more than one interpretation. Keep private provider or account information within the tool's intended access scope.

Chapter 41Reports and evidence-backed briefings #

Open Reports in the app or the report library. The system distinguishes periodic daily/weekly/monthly editions, requested special reports, local/browser research outputs and personal My Watchtower briefings.

Read an edition

1. Identify the edition date and covered period: daily, weekly or trailing calendar-month windows are not interchangeable.
2. Read the executive briefing and domain sections, then open the source/coverage register.
3. Use Inspect edition evidence where entitled. The evidence view binds to the edition's captured source identity rather than silently substituting today's data.
4. Inspect source fields/pages or download the source manifest. Preserve original observations, source state, query scope and model versions.
5. Download the offered HTML/PDF/JSON formats and writing style. The selected style should not change the underlying factual evidence or purchased scope.

An edition can include current evidence, retained/aged context, unknown clocks and unavailable sources. Its source count is not a number of independent verified incidents. A lack of capture history limits what “changed over the period” can establish. Report documents can remain available after some raw evidence retention expires; use the manifest and any explicit evidence-expiry notice.

Request and purchase deliberately

Choose the report's topic/scope and inspect availability and the displayed price before confirming. At this edition, the public site advertises special reports at US$2 each, including PDF and HTML; current checkout is authoritative. Optional research requests may have separate credit/provider costs.

After a payment or generation request, use the existing purchase/confirmation status and report library. A pending confirmation is not an instruction to pay again. Download under the purchasing account, and retain the original edition identity with your copy. Reporting release.

Chapter 42Replay, saved work and exports #

Three different kinds of history

Application replay uses a captured browser/application snapshot and the data it contains. Source history is the provider's or collector's retained dated series/records. Prospective forecast history is the immutable record of forecasts actually issued and later resolved or left awaiting evidence. These histories should not be conflated.

When you enter replay, some workspaces restore captured inputs and others deliberately withhold live views. A missing historical field should not be filled with today's reading. Return to Live explicitly before expecting current source requests or monitoring. Aviation's observation replay and forecast-ledger inspection also have their own scope and retention.

Where does a save live?

Signing into the same account on a different device does not make every browser-local feature appear there. Clearing site data can erase local research even when cloud records remain. Export important local work first and check import controls for the same feature/version. A copied view link is not a backup of private evidence or notes.

Export for another reader

Include the question, selected country/period/filter, source URL, source dates, units, missingness, model/scenario version and your interpretation. Explain whether the file is a displayed selection, a complete filtered export or a manifest. Keep source attribution/licensing notices. Snapshot conflicts should trigger a coherent reread, not silent page mixing.

Chapter 43API Connections and AI-agent MCP access #

Your optional provider connections

Open User & Settings → API Connections & Add-ons. Use the catalogue and fields shown for an eligible provider, save through the account's connection controls, then check connection/capability status. Do not assume that possessing a key grants every endpoint or redistribution right.

Stock/fundamental and aviation add-ons retain provider-specific placement and use conditions. Private credentials belong in the connection controls, not in a research prompt, URL, public note, screenshot or exported configuration. Manual private provider results remain distinct from the shared public source corpus. Provider charges and Watchtower access are separate.

Connect an AI agent

1. Sign in with an API-capable account, open Advanced and find AI Agent MCP — Hosted Agent Access.
2. Create an appropriately scoped token. The full token is shown at creation; store it in the client's secret controls.
3. Configure the client's remote HTTP MCP endpoint as https://watchtower.technology/agent/mcp, with Authorization: Bearer YOUR_WATCHTOWER_MCP_TOKEN.
4. Initialize the connection and request the live tool schemas. The documentation page /agents/ is not the MCP endpoint.
5. Start with watchtower.list_data_routes, then read a discovered path with watchtower.read_data_route.

At the guide audit, the reviewed catalogue contained 191 routes. This is a dated catalogue count, not a fixed availability or unique-observation total. The authenticated catalogue and current tool schemas govern access.

Read complete, coherent pages

An illustrative tool call for discovery is:

{"name":"watchtower.list_data_routes","arguments":{"category":"weather-hazards","query":"natural-hazards"}}

Use a returned path for the next call, for example:

{"name":"watchtower.read_data_route","arguments":{"path":"/agent/api/natural-hazards-evidence","params":{"view":"overview"},"limit":50}}

These are request examples, not current observations. Preserve snapshot, capturedAt, snapshotExpiresAt, source dates and status. If page.hasMore is true, continue with the returned page.nextOffset and the same query/snapshot. Use the returned text pointer/offset for long strings. A JSON Pointer selector can focus on one subtree.

Two pagination levels can exist: the reader's JSON-field pages and the source endpoint's own row pages. Finishing the first does not prove that you obtained all upstream records. The MCP snapshot and an endpoint's params.snapshot are also distinct. On expiry or mismatch, restart a coherent read rather than joining generations.

The seven large-data getters now use bounded paging for oversized results. Search covers the canonical retained corpus; live route reads use reviewed current handlers. Private workspaces, credentials, account/payment/admin records and mutation routes are excluded from the shared reader. Report generation through Deep Analyst is a separate, explicitly requested and separately entitled action. Full setup examples and the machine-readable manifest are in the agent guide.

Chapter 44Accounts, subscriptions, affiliates and administration #

Account and billing

Use Account / Access to verify the active identity and entitlement. Allow access checking to finish before editing private work. If a view is gated, check the account capability and source requirement rather than assuming a data failure.

At this edition, the public Pro offer is US$29/month or US$250/year, with PayPal recurring checkout on the public website. Confirm the final price, tax, provider and renewal terms shown for your account before approving. Other configured payment options, if presented, retain their own terms. Special reports and optional provider/AI usage can be separate costs.

If checkout returns pending or for another account, sign in with the purchasing account and use Refresh Access, the activation/check-payment link or the purchase's existing status. Do not infer successful entitlement from a URL alone. Contact support for unresolved billing, access, account-data or deletion requests, and consult the Privacy Notice and Terms. Financial records may have retention obligations separate from active account access.

Affiliates

Affiliate access is invitation/approval-based. An ordinary signup or a referral link does not create an approved affiliate automatically. Where enabled, complete the displayed terms and payout-identity review requirements and inspect the status before expecting commissions or payments. Approval, review, pending payout and confirmed payment are distinct stages. Current programme terms govern amounts and eligibility.

Authorised administration

Backoffice is for authorised account/access, sales and affiliate administration. Account deletion, manual access grants and financial decisions are privileged workflows; normal user access does not imply those powers. App Health provides authorised diagnostics, report status and source/runtime context, including forecast-operations monitoring. A “fixed” health report should reflect reviewed evidence, not merely a later successful page load.

Forecast-resolution/reviewer roles are separate from general administration. Matured authored forecasts require their defined resolution/evidence process; scheduled maturity alone does not justify marking them correct or incorrect. Personal pilot administration likewise requires the proper role and invitation/consent workflow.

Chapter 45Troubleshooting and reporting a problem #

A tab is missing

Check Daily Summary/Simple/Advanced mode and your current access. Look in Help, Commands, the horizontally scrollable tab strip and the expandable Research/Finance/OSINT area. Review Customize Workspace or a saved startup layout. A renamed tab can preserve its older underlying identity; for example, Global Debt contains the retained Real Estate research.

The page is loading or the map is blank

Wait for the actual Foundation/Workspace/Data stages; a loading animation is not a completion percentage. Use the supplied retry/reload control if loading fails. Open the intended tab and its map explicitly. On compact screens, close a control rail that covers the canvas. If a graphics-recovery control is offered, use it before repeatedly recreating the view. Source imagery can genuinely be unavailable at a selected place or zoom.

A closed main globe deliberately stops its own live/rendering work. Independent workspaces have separate visibility lifecycles. A source read can also be slower than the interface itself. Check the source status and original timestamps after recovery rather than assuming a new fetch created a fresh observation.

A number is absent, stale or different from another view

Check selected unit, period, valuation, currency, source, filters and snapshot. One view may show a reference catalogue, another a qualified current subset and another retained history. Negative source results, blocked access or an unknown clock must not be replaced with a guessed zero. For a multi-page snapshot conflict, refresh through the workspace and restart the coherent export.

The session says reconnecting

If Session: reconnecting, Connection interrupted or Retry connection appears, private tools are paused while the account connection is verified. Restore connectivity and use the offered retry control if needed. This differs from a confirmed sign-out or changed-account response. Check the final account identity/access state before resuming private edits; the presence of a previous label is not an active entitlement.

An alert did not arrive

Identify the monitoring system first. Was it a browser rule, server inbox, Aviation notification or a passive tracked item? Check armed state, known inputs, duration, cooldown, daily budget, timezone and the actual delivery channel. Browser permissions and page visibility matter to browser notifications. A tracked ID alone is not a notification subscription.

Saved work is missing

Check the account, device/browser, storage location and selected profile before resetting anything. Browser-local and session work will not necessarily appear on another device. Use the feature's backup/import recovery controls and retain any corrupt-preference backup offered. A refresh-access or replay transition can intentionally clear an active private view while verification runs.

Submit a useful report

Use Report Error, Help or support. Describe the workspace, action, approximate time/timezone, expected versus actual result, browser/viewport and any visible source/error identifier. A screenshot showing the affected controls can help; remove private holdings, notes and credentials. Report the source status as shown. One failed provider, a graphics issue and an access error require different remedies.

Chapter 46Practical end-to-end playbooks #

A five-minute morning brief

Open Pro Brief or Daily Summary, read the leading source-dated developments and prices, then check My Watchtower's changes/inbox and the upcoming Calendar. Choose two items for deeper review, open their source details and acknowledge the personal items you have read. Finish with a short list of what changed, what matters to your question and what remains unverified. Avoid equating the number of headlines with the number of independent events.

Compare countries for a long-term research question

Start in Countries and compare the same indicators/years. Open Global Politics for governance history, Ancap Welcomeness for its declared value-based pillars, Taxes & fees for source-qualified fiscal details, and Wealth/Global Debt for economic context. Preserve nominal/real/PPP distinctions and coverage gaps. Your conclusion should explain the criteria you chose rather than presenting a model rank as universally best for every person.

Investigate an energy or supply-chain disruption

Identify the source-dated development in Search/News or Linked Events. Inspect the relevant place in Commodity Atlas and Infrastructure, distinguishing precise geometry from schematic/locality references. Add Marine/Mobility and official weather evidence as separate context. Use Energy's explicit supply scenario only if the inputs support it, and label all assumptions. Save an investigation with supporting and contradictory evidence; do not convert a nearby asset into a measured production loss.

Research a company behind a commodity theme

Choose the commodity and source period, identify reference companies, then resolve the issuer and listing in Companies. Inspect primary financial facts and current issuer disclosures for production/cost/reserve claims. Use a market chart for its own price question. Keep country production totals, company category membership and mine/project exposure separate. Export the selected facts and record the remaining disclosure gaps.

Review a banking concern

Begin with the precise claim and its original date. Use Bank Run Monitor for institution/system evidence, Financial Stress and Deposits for broader funding context, Global Debt for relevant balance-sheet definitions and Countries for fiscal/market references. Look for independent confirmations and counter-evidence. A payment outage, a market-price decline and a deposit outflow are not the same event or proof of the same cause.

Monitor a weather-related aviation question

Check the official warning's issue/effective/expiry period in Weather Alerts. In Aviation, inspect source-native positions and contact clocks, then airport/METAR/TAF/airspace context. Use retained replay only for observations actually captured. Read original official notices; horizontal overlap does not establish flight impact. Save an appropriately scoped watch or case and identify whether you expect a server inbox item or an open-browser notification.

Evaluate a forecast instead of collecting a dramatic number

Open the target definition, base/target time, source cutoff, model version and range. Compare the baseline and relevant historical tests. Find the genuinely issued record and subsequent compatible outcome if available. Distinguish unresolved, awaiting-source and source-revised states. Explain the calibration and sample limitations before using the estimate in another report. Compare identical definitions rather than mixing country context with global or fund-price targets.

Create a reproducible briefing for another reader

State the question, period and audience. Gather a small set of source-dated facts from the relevant workspaces, record contrary evidence and gaps, and use a report or cited investigation export. Include the manifest/evidence identifiers where available, native units and all scenario assumptions. Check that the recipient can access the linked material and that the export's licence permits the intended use. Preserve a copy of the edition you actually used.

Chapter 47What is new #

This edition brings the major October updates into one handbook. The release links explain the implementation and source qualifications in more detail; the guide chapters above explain the workflow.

Chapter 48Glossary and interpretation reference #

Chapter 49Workspace directory #

This directory is generated from the deployed navigation inventory checked for this edition. It links each registered tab to the relevant explanation in this guide. Some entries are compact-panel tabs or access/layout-dependent research tools; not every entry appears in the primary strip at once.

Chapter 50Sources, methods and keeping the guide current #

Use the live data catalogue for reviewed dataset interfaces and registered fields, and Updates for releases after this edition. The guide's coverage audit is anchored to the deployed interface, current public methods and verified release records. It does not reproduce private account records or a live data feed.

Feature methods are the best place for exact formulas, source scope, retention and export qualifications. Existing pages can describe an earlier release; use the current workspace's version/date alongside them. In particular, the restored Fear & Greed/bond-unit and company-country flag releases supplement the earlier market-presentation methods, and Aviation investigations supplement the original tracking introduction.

• Bayes Threat · Freedom & Prosperity
• Energy prices, futures and production impacts
• Outcome Tree · Global Risk Studio
• Multi-input intelligence rules
• Global Health Observatory
• Market Radar
• Conditions, forecasts and outcomes
• Global Disaster Watch & Forecasts
• Geopolitical Outlook
• Forecast Lab: models, evidence and resolution
• Worldwide fibre and cable references
• Bank Run Monitor: evidence guide
• Health & Biosecurity facility references
• Worldwide markets, with explicit source meaning
• War Room workbench
• Economic Calendar: sources and methods
• Natural Hazards: source-qualified global evidence
• War Room: terrain, communications and hazard methods
• Public camera explorer
• The World Newsroom
• Companies: source-backed research
• Country taxes, fees and scoring
• Aviation investigations
• Subsea connections and mapped underwater utilities
• Ancap Welcomeness
• Space Weather Observatory

Branding and typeface references are listed in the site's image and font credits.

For account or product questions, use Help or admin@watchtower.technology. When sharing a finding, keep its source dates, original links, units, coverage and assumptions with it. That habit makes Watchtower's breadth useful rather than simply making a dashboard busier.