# Truth Bear - Verifiable Screening Intelligence for AI Agents (x402, USDC on Base, no API key; GAUGE engine) > Pay-per-call data signals for AI agents via the x402 protocol (USDC on Base, no API key, no signup). Headline value: (1) pre-joined CROSS-LINE CASCADE — one call surfaces downstream exposure across industries; (2) LEADING signals — the upstream factor before the current state; (3) aggregation / screening convenience over scattered free official sources (USGS/NOAA/EPA/SEC/FDA/EIA/…), pre-cleaned into one schema; (4) verifiable official-source provenance (official timestamp + record_hash) for those who need citable screening evidence. Pure description — no judgment, no prediction; official-source facts + back-testable statistics, the consumer decides. > Provenance is scoped, not the headline: record_hash (canonical sha256, recomputable offline) + official source_ref timestamp cover verification. NOTE — Bitcoin / OpenTimestamps anchoring was DISCONTINUED on 2026-07-23 and is not offered; do not build verifiers against it. Delivered records carry ots_status only as a retired, always-pending compatibility field. ## How to pay x402 (HTTP 402): GET an endpoint; on 402 pay the quoted USDC amount on Base (EIP-3009, gasless) with any x402 client; retry with X-PAYMENT. USDC settles directly to the provider wallet. No accounts, no API keys. Machine-readable discovery: /.well-known/x402 and /manifest. Per-channel projections of the same catalogue (pick the shape your stack parses; all three are generated from one source, so they cannot drift apart): /manifest (x402 resources with price + inputSchema + outputSchema) | /manifest?platform=mcp (MCP tool list, identical to POST /mcp tools/list) | /manifest?platform=a2a and /.well-known/agent-card.json (A2A AgentCard; note this agent does NOT implement A2A message transport - skills are plain HTTP GET endpoints paid with x402) | /.well-known/agent.json (legacy Truth Bear capability card, kept unchanged for existing consumers). IMPORTANT — raise your client's maxValue for endpoints above $0.10: the common x402 client helper (`wrapFetchWithPayment(fetch, account)`) defaults `maxValue` to 100000 atomic units = $0.10 USDC. Endpoints priced above that (the three-leg cross-verification bundles, and the subscription; see /manifest for each current price) are rejected **inside your own client before any request reaches us** — you see a failure with no server-side trace and no 402 to inspect. Pass an explicit cap — `wrapFetchWithPayment(fetch, account, 10000000n)` (= $10) — or read the `client_hint` object attached to the 402 body of those endpoints, which states the required atomic amount and the default cap. Endpoints at or below $0.10 need no change. IMPORTANT — pay from a plain EOA: the paying wallet must be a plain externally-owned account (EOA) with NO contract code. Smart-contract wallets and EIP-7702-delegated accounts verify signatures via EIP-1271 instead of ecrecover; some implementations reject the EIP-3009 authorization, so payments fail intermittently with the opaque facilitator error "invalid_payload" even though the signature, balance, and network are all correct. If you hit invalid_payload, check `eth_getCode(payer)` — a non-empty result means your wallet is delegated; pay from an un-delegated EOA instead. ## Verifiable citation (offline, zero-trust) Every paid JSON response carries a "_citation" — a signed, offline-verifiable fact envelope: Ed25519 signature over RFC-8785/JCS canonical bytes, record_hash = sha256, statement_type=description / not_advice=true (pure description, never advice/prediction). Recompute + verify it with the standalone verifier (citation-object/verify.mjs) — zero network, no trust in this endpoint required; tampering with any factual field breaks both hash and signature. Issuer identity is an offline did:key (also in /manifest.citation and /.well-known/agent.json). Issuer DID: did:key:z6MkjSqVEigLT2eyFzJ9Vhb3DJD2wmwGosSbk1PJ1GTnuABV (kid=kid-2026-07). MCP (Model Context Protocol): POST /mcp speaks JSON-RPC 2.0 over HTTP (initialize / tools/list / tools/call; no SSE). Two FREE tools — verify_citation (hand it any Truth Bear record_hash; it re-looks-up and recomputes the canonical hash, and tells you in plain language exactly what that hash attests) and find_signal (coverage + freshness manifest, so you can check is-my-entity-covered-and-how-fresh before paying). One payment-pointer tool — get_official_record returns the REAL x402 challenge (accepts[]: network / asset / payTo / amount) for a signal_id+entity; this MCP server never takes payment and never returns paid data, you pay at that URL with your own wallet and receive the record directly. Add it to any MCP host as an HTTP/streamable server pointing at /mcp. Channels (x402 is one channel, not the only one — see GET /gauge/channels): x402 (AI-agent micropayment), a stateless subscription token (no account needed; /gauge/subscribe -> Bearer token for /gauge/watch, /gauge/watch-changes, /gauge/manifest-sweep), and - where the operator configures it - a fiat/card rail returning the same JSON (~2% fee; the system never handles the payment secret). RapidAPI / B2B-direct / academic (Zenodo+DOI) are roadmap channels. Every paid JSON response carries a machine-readable "grade: screening-level, not decision-grade" + "scope" (monitoring-not-forecast, intelligence-not-command), a "provenance.data_lineage[]" (per-signal source + independent_dimension + uncertainty; uncertainty says "not-provided-in-this-projection" when a projection did not carry the multi-source agreement CV, rather than asserting single-source), "decision_points" (one-tag-swap = one what-if call), and — once a prior call outcome is on record — a "last_fact" (the single objective later official reading; not a prediction/average/hit-rate). ## GAUGE — environmental signals (hydrology × air quality × precipitation) Flood-risk & river-anomaly check: real-time US river water level (gage height), streamflow (cfs), flood stage / flood category vs official USGS/NOAA thresholds, plus 5-year seasonal statistical anomaly (percentile / strata) and record_hash verifiability. Two rulers per signal (official thresholds + statistical distribution); cross-line corroboration (rain↔river↔air) as material for the consumer to self-verify causal chains. - Data-acquisition contract: GAUGE serves screening records from STORED snapshots (SQL), never a live upstream call at request time — every paid response that carries a timestamp also carries provenance.data_as_of + provenance.age_hours_now (age computed at request time, not frozen at collection) + provenance.freshness + provenance.freshness_basis (the exact cadence and thresholds used for that label). Freshness is CADENCE-AWARE and uses the SAME function as the free /gauge/coverage manifest, so the label you see before paying is the label you get after paying: monthly/quarterly/annual official lines are not mislabeled stale by a 24h rule, and a line whose upstream froze is labeled stale + dead_source regardless of when it was last collected. Per-cadence thresholds are machine-readable at /gauge/coverage (freshness_thresholds). Real-time-critical lines are collected hourly, others daily/weekly. Stored-serving avoids upstream API latency; use each record within its data cadence. - Free: /gauge/coverage (coverage + freshness manifest — is-my-entity-covered-and-how-fresh before paying; ?industry=&signal_id=&entity=&summary=1) · /gauge/history-preview?signal_id=&entity= (last 7 snapshots, time-series shape) · /gauge/raw?signal_id=hydrology.river-level&entity= (base reading) · /gauge/sample · /gauge/preview · /gauge/catalog - Skill (paid): /gauge?signal_id=&entity= — full flood-risk/anomaly record (price: see /manifest) - Add-ons (single rulers, cheapest tier — see /manifest for each price): /gauge/ruler · /gauge/calibrated · /gauge/strata · /gauge/history (seasonal percentile/frequency) · /gauge/air?loc= · /gauge/precip?loc= - History time-series (paid): /gauge/history-series?signal_id=&entity=&days=&format= — full retained time-series (price: see /manifest) (snapshot_date ASC, up to days default 400 cap 1000; columns snapshot_date/observed_at/data_age_hours/current/band/record_hash), format=csv or json, for research/actuarial/base-rate. NOTE: history depth grows over time (data is being accumulated; early records may be sparse); curated/emerging lines snapshot weekly while live lines snapshot daily/hourly, so time-series granularity varies by line. - Bundle: /gauge/region?loc= (air + precipitation + nearby river, one location) · /gauge/census?signal_id= (cross-station scan). Prices: see /manifest - Signals: hydrology.river-level, hydrology.streamflow, airquality.aqi, airquality.pm25, precipitation.daily, precipitation.wetness30d - Sources: USGS Water Services, NOAA NWS/NWPS, US EPA AQI, Copernicus CAMS, ECMWF ERA5 ## Subscription / watchlist (stateless one-time token) Pay ONCE at /gauge/subscribe (x402) to mint a STATELESS signed token (HMAC; no account, no login, no session, no server-side database). Save the token and send it as Authorization: Bearer or ?token= on the two subscription-gated endpoints below — no per-call payment. Renew by calling /gauge/subscribe again. The client holds its own watchlist; nothing is stored server-side. The FREE tier (coverage, history-preview, discovery, single-call region pay-per-call) is completely unchanged. If the operator has not set GAUGE_SUB_SECRET, subscription endpoints return a not-configured message and free/paid routes are unaffected. - Register (paid): /gauge/subscribe — one-time x402 payment mints { token, plan: watchlist, expires_at, days } valid for GAUGE_SUB_DAYS (default 30). No account, no login. - Batch read (free, token-gated): /gauge/watch?items=industry:entity,industry:entity,... — one call returns the whole watchlist state (up to 30 industry:entity items), each with its latest per-signal bundle (signal_id, entity_name, current, band, observed_at, data_age_hours, record_hash) + provenance, served from stored SQL. Example ?items=drug-detail:carboplatin,border-wait:laredo-world-trade,drug-shortage:oncology. Invalid/expired/missing token → 401. - Poll deltas (free, token-gated): /gauge/watch-changes?items=...&since= — for each item returns only signals whose observed_at is newer than ?since (default 24h ago). Webhook-lite monitoring; a true push webhook is a future addition. ## Agriculture triangle — crop drought / heat / vegetation health (global grain belts) Yield & drought risk for AI agents (ag traders, crop insurers). Three co-located legs per grain region + cross-validation (has drought hit the crop canopy yet). Official sources, verifiable. - Skill: /gauge/agri-region?loc= — $0.25, one region: agricultural drought + heat + crop VHI + cross-check - Signals (query via /gauge, /gauge/ruler, /gauge/strata, /gauge/history): agriculture.drought (soil moisture 28–100cm vs official US Drought Monitor D0–D4), agriculture.heat (Tmax heat-stress + growing degree days), agriculture.crop-vhi (NOAA satellite Vegetation Health Index, VHI<40 = crop stress) - Sources: ECMWF ERA5 (soil/temperature), US Drought Monitor (D0–D4 classification), NOAA STAR VHP (satellite vegetation health, 1982+) - Grain regions: US Corn Belt (Iowa/Illinois/Nebraska/Kansas), Argentina Pampas, Brazil Mato Grosso, Ukraine, Russia (Krasnodar), China (N. China Plain), Australia NSW, India Punjab ## Power grid triangle — demand stress / renewable resource / energy inflation (global grids) Grid stress, power price & energy inflation signals for AI agents (power/energy traders, macro/inflation traders, utilities, renewable investors). Three legs per grid region + cross-validation (high demand × low renewables = grid squeeze → energy price/inflation up). Official sources, verifiable. - Skill: /gauge/grid-region?loc= — $0.25, one grid: electricity demand + renewable resource + energy inflation/gas price + cross-check - Signals (query via /gauge, /gauge/ruler, /gauge/strata, /gauge/history): electricity.demand (temperature degree days HDD+CDD → grid load pressure proxy), electricity.renewable (solar irradiance + wind → renewable generation potential), energy.inflation (US CPI Energy YoY, official inflation), energy.gas-price (Henry Hub natural gas spot, power-price driver) - Sources: ECMWF ERA5 (temperature/solar/wind, weather-driven proxy — not actual MW), FRED (US BLS CPI Energy CPIENGSL, EIA Henry Hub DHHNGSP) - Grid regions: US ISOs (ERCOT/CAISO/PJM/MISO/NYISO), Germany, France, UK, Spain, Japan, Australia NEM, India. Energy inflation/gas price are US-national. ## Shipping / port logistics triangle — route disruption / sea-state / vessel throughput (global chokepoints) Global logistics flow-rate signals for AI agents (shipping lines, commodity/freight traders, forwarders, ports, marine insurers). Three legs per chokepoint + cross-validation (has the sea-state disruption actually reduced vessel flow). Official + AIS sources, verifiable. - Skill: /gauge/shipping-region?loc= — $0.25, one chokepoint: logistics disruption + sea-state cause + AIS throughput + cross-check - Signals (query via /gauge, /gauge/ruler, /gauge/strata, /gauge/history): shipping.route-disruption (sea state WMO + wind vs operational thresholds → operable/marginal/disrupted/severe, logistics-flow proxy), shipping.sea-state (significant wave height Hs, WMO sea state 0–9 + wind-wave/swell decomposition → local-storm vs distant-swell origin), shipping.throughput (live AIS vessel count near port: waiting/anchored vs transiting + congestion, key-gated) - Sources: Open-Meteo Marine (ECMWF wave: Hs/wind-wave/swell/period, weather-driven proxy — not actual MW/throughput), ERA5 (marine wind), aisstream.io (live AIS, rate-limited sample; coastal-good/open-ocean-sparse) - Chokepoints: Suez/Red Sea, Panama, Malacca, Hormuz, Cape of Good Hope, English Channel, Biscay, North Sea, Gulf of Mexico, South China Sea, Bay of Bengal, North Pacific. ## Regulatory filing triangle — corporate disclosure / distress early-warning (SEC EDGAR) Corporate distress early-warning signals for AI agents (event-driven/activist/short funds, quant funds, credit & distressed analysts, forensic auditors, compliance). Three legs per company + cross-validation (do insiders confirm or contradict the disclosed distress). All official SEC EDGAR, verifiable. - Skill: /gauge/filing-company?entity= — $0.25, one company: 8-K material events + NT delinquency + Form 4 insider buy/sell + cross-check - Signals (query via /gauge, /gauge/census): filing.material-events (8-K frequency + item severity: bankruptcy 1.03 / default 2.04 / restatement 4.02 / delisting 3.01 = critical; impairment/termination = high; exec change 5.02 = elevated; + vs own historical baseline percentile), filing.delinquency (NT 10-K/10-Q late-filing notifications = distress leading indicator), filing.insider (Form 4 net open-market buy/sell by parsing transaction codes P/S = downstream confirmation) - Sources: SEC EDGAR submissions API (8-K items, NT filings, Form 4) + Form 4 XML (open-market transaction codes). Official, keyless, verifiable. - Watchlist: 18 large-cap + volatile names (AAPL/MSFT/NVDA/TSLA/AMZN/META/GOOGL/BA/F/INTC/GME/AMC/CVNA/PLTR/NFLX/DIS/JPM/BAC); any CIK expandable. ## Climate policy triangle — policy lifecycle / distress / carbon-project cascade (emerging industry) Climate policy lifecycle + carbon-market exposure for AI agents (carbon traders, project developers, ESG funds, climate reinsurers, policy consultants). Complete industry cycle (birth→death) — unique to emerging industries. Three legs per policy + cross-validation (has a policy's death actually hit dependent carbon projects). Official legislation records, verifiable. - Skill: /gauge/policy-region?entity= — $0.25, one policy: lifecycle + distress events + dependency cascade + cross-check - Signals (query via /gauge, /gauge/census): climate-policy.lifecycle (enacted→repealed/expired + lifespan_days + death_type, Block A industry-development snapshot), climate-policy.distress-events (official repeal bills / court challenges / budget cuts = upstream leading indicator), climate-policy.cascade (policy death → dependent carbon projects/CDM exposed = downstream E→A cross-validation), climate-policy.industry-timeline (complete industry cycle: births/deaths per year + lifespan distribution) - Sources: official legislation records (UNFCCC, EU Commission, Congress.gov, national gazettes). Pure description of official enacted/repealed facts, no judgment/prediction. - Policies: 23 tracked (EU ETS, IRA, Kyoto Protocol, Australia carbon pricing, US Clean Power Plan, California cap-and-trade, CBAM, China ETS, RGGI, Sweden carbon tax, …); operating/ended/dead. ## AI tool / MCP triangle — alive oracle / decay / zombie (emerging industry) AI tool & MCP server aliveness + death forensics for AI agents (agent builders check a tool is alive before recommending, enterprise AI, app developers, MCP hosts). Real DNS/HTTP liveness for web tools, GitHub archived/last-push for MCP servers. Three legs per tool + cross-validation (is a dead tool still recommended). Observable facts, verifiable. - Skill: /gauge/tool-region?entity= — $0.25, one tool/MCP: alive+lifecycle + decay warning + zombie detection + cross-check - Signals (query via /gauge, /gauge/census): ai-tool.lifecycle (alive oracle + launch→death + lifespan; web-tool via DNS, mcp-server via GitHub), ai-tool.distress (page gone / pricing hike / GitHub archived / consecutive absences = upstream leading), ai-tool.zombie (dead tool still recommended by AI engines/directories = downstream cross-validation), ai-tool.industry-timeline (complete industry cycle; tools die fast, ~2yr median) - Sources: DNS/HTTP liveness, GitHub API (stars/archived/last-push), AI-tool directories. Observable facts, no judgment/prediction. - Tracked: 21 web tools + MCP servers (Jasper, Perplexity, Midjourney, Cursor, HuggingFace, MCP SDKs/servers, …); operating/ended. ## AI dependency triangle — deprecation lifecycle / notices / D→B cascade (emerging industry) AI model/API/SDK deprecation + migration-risk for AI agents (app developers, platform teams, AI tool vendors, enterprise AI). AI models die fast (~1.6yr median lifespan). Three legs per dependency + cross-validation (has a model's deprecation exposed the tools built on it). Official deprecation pages, verifiable. - Skill: /gauge/dependency-region?entity= — $0.25, one dependency: lifecycle + deprecation notices + D→B cascade + cross-check - Signals (query via /gauge, /gauge/census): ai-dependency.lifecycle (launch→deprecation + lifespan), ai-dependency.deprecation-notices (announcements / sunset dates / migration guides = strong upstream leading, AI deprecations are pre-announced), ai-dependency.cascade (model deprecation → dependent AI tools exposed = downstream D→B cross-validation), ai-dependency.industry-timeline (complete industry cycle; fast model deaths) - Sources: official deprecation pages (platform.openai.com/docs/deprecations, docs.anthropic.com model-deprecations, ai.google.dev, learn.microsoft.com Azure). Observable/official facts, no judgment/prediction. - Tracked: 20 models/APIs (OpenAI GPT-3/Codex/vision, Anthropic Claude 1/2, Google PaLM, Azure OpenAI, + operating gpt-4o/claude-3.5/gemini-1.5/llama-3). ## Promise triangle — promise-vs-delivery reconciliation (emerging industry) AI/robotics/hardware promise accountability for AI agents (due-diligence, VC/growth investors, procurement, competitive analysts). Companies/products that promised big and either shipped, slipped, or died. Three legs per committer + cross-validation (has the slippage become an unmet promise, confirmed by the delivery gap). Complete industry cycle (unique to emerging industries). Official promise/roadmap pages + delivery records, verifiable. - Skill: /gauge/promise-region?entity= — $0.25, one committer: lifecycle + promise slippage + delivery reconciliation + cross-check - Signals (query via /gauge, /gauge/census): promise.lifecycle (survival → death/acquisition + lifespan), promise.slippage (timeline delays / spec walk-backs / promise-page removals = upstream leading), promise.delivery-recon (promised vs actually-shipped gap % = downstream core signal), promise.industry-timeline (complete industry cycle) - Sources: official promise/roadmap pages, product pages, press announcements, shipment/delivery records. Observable facts, gap self-evidenced by the numbers, no judgment/prediction. - Tracked: 17 committers (Humane AI Pin, Tesla Optimus, Figure AI, 1X, Sanctuary AI, Rabbit R1, Inflection, Adept, Character.AI, Cognition/Devin, Imbue, CoreWeave, Lambda, Stability AI, Cruise robotaxi, Nikola, Magic Leap); operating/ended. ## Carbon triangle — carbon-credit project integrity & repricing (emerging industry) Voluntary-carbon-market (VCM) project integrity for AI agents (carbon buyers, ESG funds, offset due-diligence, carbon traders, project developers, climate reinsurers). Carbon projects/methodologies that were credited, downgraded, suspended, or delisted. Three legs per project + cross-validation (has the rating drift become a market repricing — price collapse/buyer flight confirming the integrity flaw). Complete VCM cycle (unique to emerging industries). Official registry/standard-body status + rating + market records, verifiable. - Skill: /gauge/carbon-region?entity= — $0.25, one project: lifecycle + rating drift + price integrity + cross-check - Signals (query via /gauge, /gauge/census): carbon-project.lifecycle (registration → registry suspension/delisting/termination + lifespan), carbon-project.rating-drift (rating-agency downgrades + standard-body reviews/methodology suspensions = upstream leading), carbon-project.price-integrity (credit price collapse / buyer flight / corporate cancellation = downstream cross-validation), carbon-project.industry-timeline (complete VCM cycle) - Sources: Verra registry, Gold Standard, CDM/UNFCCC, carbon rating agencies (BeZero/Sylvera/Renoster), reputable investigations, market repricing records. Observable/official facts, no judgment/prediction — the market numbers self-evidence integrity. - Tracked: ~19 projects/methodologies (Kariba REDD+ VCS902, Rimba Raya VCS674, Southern Cardamom VCS1748, Cordillera Azul VCS985, Katingan Mentaya VCS1477, Verra REDD+ methodologies VM0007/09/15, CDM wind-down, Gold Standard/ACR/CAR projects); operating/at_risk/ended/dead. - NOTE: this is the unified-table carbon triangle (/gauge/carbon-region). The legacy /carbon/* module (checkup/graveyard/ratings-drift) remains for existing buyers. ## Stablecoin triangle — depeg / collapse forensics (emerging industry) Stablecoin depeg & collapse forensics for AI agents (treasury/risk teams, exchanges, DeFi protocols, auditors, due-diligence). Stablecoins that collapsed, depegged, wound down, or hold peg. Three legs per coin + cross-validation (has the peg distress become a market exodus — market-cap collapse/redemption outflow confirming the depeg). Distinguishes market-depeg death from orderly regulatory wind-down with intact peg. Complete cycle (unique to emerging industries). Official/exchange/regulator records, verifiable. NOT investment advice. - Skill: /gauge/stablecoin-region?entity= — $0.25, one coin: lifecycle + peg distress + redemption flight + cross-check - Signals (query via /gauge, /gauge/census): stablecoin.lifecycle (launch → depeg-death/wind-down + lifespan), stablecoin.peg-distress (off-peg episodes / max deviation / reserve-regulatory warnings = upstream leading), stablecoin.redemption-flight (market-cap collapse / redemption outflow / delisting = downstream cross-validation), stablecoin.industry-timeline (complete cycle by mechanism) - Sources: official/issuer sites, exchange announcements, regulators (NYDFS etc.), market-cap/supply/depeg records. Observable/official facts, no judgment/prediction, no investment advice — market numbers self-evidence collapse. - Tracked: ~22 stablecoins (TerraUSD/UST, Iron/TITAN, Basis Cash, ESD/DSD, DEI, Neutrino USDN, BUSD, Fei, HUSD, Diem, USDD, TUSD, USDC, USDe, USDT, DAI, FRAX, PYUSD, FDUSD); operating/at_risk/ended/dead by mechanism (algorithmic/fiat/crypto-collateralized/hybrid). - NOTE: this unified-table stablecoin triangle (/gauge/stablecoin-region) is the descriptor form. The legacy /stablecoin-deaths/* and /trust/* modules remain for existing buyers. ## AI-compute / neocloud triangle — GPU-cloud provider death & solvency (emerging industry) AI-compute (GPU-as-a-service / neocloud) provider death forensics for AI agents (AI startups/labs with prepaid clusters, compute-orchestration/routing agents, AI-infra VCs & compute-debt lenders doing counterparty due-diligence). GPU-cloud providers that fold, get acquihired, or silently stop honoring reserved capacity — stranding prepaid credits and training runs. Three legs per provider + cross-validation (has the financing distress become a real death that stranded downstream workloads). Complete cycle (emerging industry). MAX cross-line cascade: a dead neocloud exposes the AI tools/models/promises hosted on it. Moat: a dead provider's true capacity/prepaid-obligation ledger vanishes and is un-reconstructable. - Skill: /gauge/aicompute-region?entity= — $0.25, one provider: lifecycle + financing distress + workload exposure + cross-check - Signals (query via /gauge, /gauge/census): ai-compute.lifecycle (founding → cease/acquihire/insolvency + lifespan + tier), ai-compute.financing-distress (layoffs/WARN, data-center lease default, funding stall, GPU-debt = upstream leading; capital structure dies before the API), ai-compute.workload-exposure (which hosted ai-tools/ai-deps/promises are exposed when it dies + bankruptcy-docket/DNS/domain corroboration = downstream cross-validation & cross-line cascade), ai-compute.industry-timeline (complete cycle by tier) - Sources: provider status pages / API-DNS liveness / company registries + WARN layoff notices + data-center lease filings + court/bankruptcy dockets (CourtListener/PACER) + hosted-on dependency map. Observable/official facts, no judgment/prediction. Distinguishes graceful acquihire (workloads migrated) from a stranding death. - Tracked: ~20 providers (CoreWeave, Lambda, Crusoe, Nebius, Together AI, RunPod, Vast.ai, Voltage Park, FluidStack, DigitalOcean/Paperspace, Vultr, Hyperstack, Latitude.sh, SF Compute, IREN + at-risk long-tail resellers); operating/at_risk/ended/dead by tier (hyperscaler-neocloud/mid/gpu-reseller-longtail). ## DePIN triangle — decentralized-physical-infra network death / ghost-network forensics (emerging industry) DePIN (Decentralized Physical Infrastructure Network) death forensics for AI agents (infra-supply buyers routing wireless/compute/mapping/storage workloads onto a DePIN, DePIN token/index allocators, DePIN-treasury/DAO agents). Detects the GHOST DePIN: the real-world node fleet + paid usage collapses to ~zero while the TOKEN STILL TRADES — token price is a lagging, misleading signal. Three legs per network + cross-validation (has the reward-claim churn become a real death confirmed by paid-usage flatline and downstream supply exposure). Complete cycle (emerging industry). MAX cross-line cascade: compute/storage DePIN death exposes ai-compute/ai-tools consumers; sensor/mapping DePIN death starves data-consumer lines. Moat: node-liveness + paid-usage vanish once a network goes dark, un-reconstructable. - Skill: /gauge/depin-region?entity= — $0.25, one network: lifecycle + reward churn + downstream exposure + cross-check - Signals (query via /gauge, /gauge/census): depin.lifecycle (active-node-count/paid-supply based, NOT price; operating→at_risk→ended→dead/ghost), depin.reward-churn (node operators stop claiming rewards = earliest on-chain economic unplugging, upstream leading), depin.downstream-exposure (ghost → stranded downstream consumers; compute/storage→ai-compute/ai-tools, sensor/mapping→data lines; gated by paid-usage-flatline = downstream cross-validation & cascade), depin.industry-timeline (complete cycle by vertical) - Sources: on-chain reward-distribution contracts + node registries/explorers + on-chain service-payment/data-credit/job burns (paid usage) + downstream-consumer linkage. Observable/on-chain facts, no judgment/prediction. Price is NOT the liveness signal. - Tracked: ~20 networks (Helium, Helium Mobile, Render, io.net, Akash, Filecoin, Arweave, Grass, Hivemapper, DIMO, WeatherXM, GEODNET, Nodle, Theta, NATIX, Wingbits + at-risk PlanetWatch/Pollen Mobile/MXC/Streamr); operating/at_risk/ended/dead by vertical (wireless-iot/compute/storage/sensor-mapping/energy/connectivity). ## ESG-fund triangle — greenwashing label-death & sustainable-ETF delisting (emerging industry) ESG / sustainable fund & ETF greenwashing forensics for AI agents (allocation/robo-advisor agents, ESG-screening/compliance bots, wealth due-diligence agents pre-trade, greenwashing journalists). Funds whose SUSTAINABILITY CLAIM dies — SFDR Article 9→8 downgrade, "ESG/sustainable/climate" stripped from the fund name, removal from ESG index/rating lists — or funds liquidated/delisted. Three legs per fund + cross-validation. Distinguishes LABEL death (claim died, fund keeps trading) from FUND death (liquidated). Complete cycle (emerging industry). MAX bidirectional cross-line cascade: a fund's ESG thesis resting on dying carbon-credit projects (carbon line) or repealed climate legislation (climate-policy line). Moat: funds silently re-label/re-classify in place — a hash-anchored snapshot of "what it claimed on date X" is un-reconstructable after the edit. NOT investment advice. - Skill: /gauge/esgfund-region?entity= — $0.25, one fund: lifecycle + rating drift + holdings cascade + cross-check - Signals (query via /gauge, /gauge/census): esg-fund.lifecycle (label death: SFDR 9→8/name-strip/rating-removed [fund lives] vs fund death: liquidated/delisted), esg-fund.rating-drift (MSCI ESG downgrade / Morningstar sustainable-list removal / index exclusion = upstream leading, jurisdiction-agnostic EU-SFDR/US-SEC/UK-SDR), esg-fund.holdings-cascade (holds carbon projects [carbon] or rests on climate policy [climate-policy] that are dying → bidirectional cascade, gated by official corroboration = downstream cross-validation), esg-fund.industry-timeline (complete cycle by jurisdiction & death-type; signature 2022–2023 SFDR Art9→8 wave) - Sources: SFDR classification filings + prospectus/KIID + fund/national registers (CSSF/BaFin/AMF/FCA) + exchange delisting notices + MSCI/Morningstar ESG pages + regulator greenwashing enforcement (SEC/ESMA/BaFin/FCA). Observable/official facts, no judgment/prediction, no investment advice. - Tracked: ~20 funds/cohorts (BlackRock/Amundi/PIMCO/NN/Invesco/HSBC/Handelsbanken Art9→8 downgrades, DWS greenwashing order, ESMA name-strip cohort, Morningstar delist cohort, UK SDR cohort + operating iShares ESGU/SPDR EFIV/JPM Carbon Transition); operating/at_risk/ended(label death)/dead(fund death) by jurisdiction. ## RWA triangle — tokenized real-world-asset default & redemption forensics (emerging industry) RWA (tokenized real-world asset) death forensics for AI agents (DeFi allocators & RWA-fund LPs, on-chain treasury/credit due-diligence agents auto-allocating stablecoins into tokenized yield, protocols/stablecoins holding RWA as reserves/collateral, auditors). Tokenized off-chain assets — treasuries, private credit, real estate, commodities — that DEFAULT, wind down, freeze redemptions, or whose off-chain backing evaporates while the token persists. Three legs per issuer/pool + cross-validation. Complete cycle (emerging industry). MAX cross-line cascade: a defaulted RWA held as reserves by a yield/reserve-backed stablecoin degrades that coin's backing → flags it at_risk on the stablecoin line. Moat: after a private-credit pool defaults, its true off-chain collateral state vanishes; the on-chain pre-death TVL/redemption/attestation time-series must be captured before collapse and is un-reconstructable after. Boundary: RWA = tokenized off-chain assets (failure = off-chain default / redemption freeze), distinct from stablecoin peg-break; a stablecoin holding RWA is the cascade edge, not an overlap. NOT investment advice. - Skill: /gauge/rwa-region?entity= — $0.25, one issuer/pool: lifecycle + redemption stress + exposure cascade + cross-check - Signals (query via /gauge, /gauge/census): rwa.lifecycle (launch → default/wind-down/redemption-frozen/delisting; by asset-class), rwa.redemption-stress (redemption pause/gating / queue lengthening / NAV discount = upstream leading, fully on-chain across all asset classes, most reliable death precursor), rwa.exposure-cascade (on-chain default/writedown + which stablecoins/lending-markets/vaults hold the dead RWA as reserves/collateral → cascade into stablecoin line; gated by on-chain default corroboration = downstream cross-validation), rwa.industry-timeline (complete cycle by asset-class; private-credit 2022–23 default wave) - Sources: on-chain pool/vault state (TVL/NAV/redemption-queue via public RPC/subgraph) + issuer attestations/audits + rwa.xyz/DefiLlama/Dune + on-chain default/writedown events + downstream reserve/collateral holdings graph. Observable/on-chain/official facts, no judgment/prediction, no investment advice. - Tracked: ~20 issuers/pools (BlackRock BUIDL, Ondo OUSG/USDY, Franklin BENJI, Superstate, Hashnote USYC, M^0, Backed, Securitize, Maple Finance + defaulted Orthogonal/Auros pools, Goldfinch, TrueFi, Centrifuge, Figure/Provenance, Clearpool, PAXG/XAUt gold, RealT); operating/at_risk/ended/dead by asset-class (tokenized-treasury/private-credit/real-estate/commodity/invoice-receivable). ## Smallsat triangle — constellation death / deorbit forensics (emerging industry) Smallsat constellation death forensics for AI agents (satellite-data procurement agents / imagery-IoT-backhaul buyers, space-asset insurers/reinsurers, satellite-IoT & imagery integrators doing supplier DD, LPs/acquirers). Operators whose constellation dies — fleet degrades (N-of-M birds alive), operator goes bankrupt/winds down, spectrum license lapses, or satellites deorbit/re-enter. Three legs per operator + cross-validation. Complete cycle (emerging industry). Distinguishes CORPORATE death with a still-flying fleet (OneWeb→Eutelsat) from a real fleet deorbit. The third leg is un-fakeable orbital physics. MAX gated cross-line cascade: a dead EO/weather/comms constellation strands downstream data consumers → agriculture/shipping/precipitation/airquality/depin/ai-deps. Moat: after an operator abandons its dashboard, "how many birds were actually flying" is un-reconstructable from the open web — only the orbital catalog survives. - Skill: /gauge/smallsat-region?entity= — $0.25, one operator: lifecycle + spectrum milestone + orbital decay + cross-check - Signals (query via /gauge, /gauge/census): smallsat.lifecycle (deploying → operating → degraded → wound-down/acquired vs bankrupt/deorbited; by constellation-type), smallsat.spectrum-milestone (FCC/ITU bring-into-use / deployment-milestone miss / license cancellation = upstream leading, leads fleet death by months-to-years), smallsat.orbital-decay (Space-Track/CelesTrak TLE active-ratio vs claimed / loss of station-keeping / deorbit = downstream cross-validation, physically un-fakeable; gated cascade into agriculture/shipping/precipitation/airquality/depin/ai-deps), smallsat.industry-timeline (complete cycle by constellation-type) - Sources: SEC/EDGAR + bankruptcy dockets (PACER) + FCC IBFS / ITU spectrum filings + Space-Track.org SATCAT & CelesTrak TLE (active-count/decay/re-entry). Observable/official/orbital facts, no judgment/prediction. - Tracked: ~19 operators (Planet Labs, Spire, ICEYE, BlackSky, HawkEye 360, Kepler, Fleet Space, Starlink, Astranis, Loft Orbital + OneWeb/Swarm ended + LeoSat/Sky-and-Space/Audacy dead + at-risk Satellogic/Capella/Astrocast/Momentus); operating/at_risk/ended/dead by constellation-type. Launch-only vehicle companies belong in the promises line. ## LRT triangle — liquid-restaking-token depeg & slashing forensics (emerging industry) LRT (liquid restaking token) & restaking-protocol distress forensics for AI agents (restakers & DeFi money-markets using LRTs as collateral → liquidations on depeg, LRT-allocator yield agents, risk agents managing leveraged LRT positions). LRTs (weETH/ezETH/rsETH/pufETH/rswETH) and restaking layers (EigenLayer/Symbiotic/Karak AVSs) that suffer a redemption/depeg death-spiral or a slashing/exploit permanently impairing the receipt token. Three legs per LRT + cross-validation. Complete cycle (emerging industry). Failure mode = redemption-queue length + slashing + discount-to-ETH-NAV — the peg is ETH-equivalent yield value, NOT a $1 fiat peg (distinct from the stablecoin line). MAX cross-line cascade: an LRT death propagates to yield/CDP stablecoins collateralized by it and to DeFi lending markets that listed it. Moat: the pre-death withdrawal-queue depth, discount curve, and operator-delegation graph are overwritten as the queue drains and are un-reconstructable after collapse. - Skill: /gauge/lrt-region?entity= — $0.25, one LRT: lifecycle + redemption queue + collateral cascade + cross-check - Signals (query via /gauge, /gauge/census): lrt.lifecycle (operating → at_risk → impaired/ended → dead; keyed on TVL + redemption-queue + peg-to-ETH-NAV, by restaking layer), lrt.redemption-queue (withdrawal-queue lengthening = upstream leading, lengthens before the discount blows out, earliest structural & most un-reconstructable signal), lrt.collateral-cascade (LRT-as-collateral de-listing/liquidation in Aave/Morpho/Euler → cascade into stablecoin/defi-lending; gated by on-chain slashing/exploit corroboration = downstream cross-validation), lrt.industry-timeline (complete cycle by restaking layer) - Sources: on-chain TVL + EigenLayer/Symbiotic/Karak withdrawal-queue contract state + DEX/CEX LRT-vs-ETH-NAV discount + on-chain slashing/exploit events + DeFi lending-market collateral risk-params/liquidations. Observable/on-chain facts, no judgment/prediction, no investment advice. - Tracked: ~19 LRTs/protocols (Ether.fi weETH, Renzo ezETH [Apr-2024 depeg], Kelp rsETH, Puffer pufETH, Swell rswETH, Eigenpie, Mellow, Bedrock, EigenLayer, Symbiotic, Karak, Restake Finance, Inception, Rio, Pendle LRT markets, ClayStack, Affine, Genesis, Stader); operating/at_risk by restaking layer (eigenlayer/symbiotic/karak). Peg is ETH-NAV not $1. ## Wildfire causal-chain (mature real-time monitoring, not a death registry) Real-time wildfire causal chain for AI agents (property/wildfire insurers, electric utilities doing PSPS de-energization, rail/road logistics, ag). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): fire-weather (hot+dry+windy) → active-fire (satellite hotspots) → downwind smoke (PM2.5). Confidence is read by how many stages are lit in causal order — fire-weather red-flag with no fire = a high-risk warning; all three lit = an active event impacting air. This exact moment's satellite + weather values an LLM cannot know live. - Skill: /gauge/wildfire-region?entity= — $0.25, one region: fire-weather + active-fire + smoke-impact + sequential-confirmation - Signals (query via /gauge, /gauge/census): wildfire.fire-weather (temp+RH+wind red-flag index = upstream leading, Open-Meteo), wildfire.active-fire (NASA FIRMS VIIRS active-fire hotspot count + FRP = current, needs free FIRMS map key), wildfire.impact-confirm (downwind PM2.5 = downstream cross-validation, Open-Meteo) - Sources: NASA FIRMS VIIRS (satellite thermal-anomaly) + Open-Meteo weather + Open-Meteo air-quality. Observable/official facts, screening-level, no judgment/prediction. Satellite hotspots include non-wildfire thermal anomalies; PM2.5 is a cross-validator not a sole attribution. - Tracked: 12 fire regions (US California/PNW/Southwest/Rockies, Canada BC/Alberta, Australia NSW/Victoria, Mediterranean Greece/Iberia, Amazon, Siberia). NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. ## NAS air-transport ground-stop triangle (mature real-time monitoring, not a death registry) Real-time FAA NAS (National Airspace System) operational-disruption causal chain for AI agents (air-cargo/air-freight operators, logistics/supply-chain planners, travel agents). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): aviation-weather (METAR ceiling/visibility/convective/gust, leading upstream) → ground-ops (current FAA state — ground stop / GDP / AFP / delay / closure, with reason_category) → impact-cascade (realized delay + network propagation to downstream hubs + air-cargo exposure). Confidence is read by how many stages are lit in causal order — adverse aviation weather leading with no FAA program = a high-risk warning; all three lit = an FAA program declared with realized delay propagating downstream. This exact moment's FAA NAS + METAR values an LLM cannot know live. This is AIR-transport identity — shipping = maritime line; the air-cargo tie to maritime is a pointer only. Live-fetch, free no-key sources. - Skill: /gauge/nas-region?entity= — $0.25, one airport: aviation-weather + ground-ops + impact-cascade + sequential-confirmation - Signals (query via /gauge, /gauge/census): nas.ground_ops (current FAA state + reason + reason_category WEATHER/VOLUME/EQUIPMENT/RUNWAY/OTHER + avg_delay_min + airport_role + cargo_hub = current, band normal/minor_delay/afp/gdp/ground_stop/closure), nas.aviation_wx (METAR flight_category + ceiling_ft + vis_sm + gust_kt + convective severity = upstream leading, band vfr_clear/mvfr/ifr/lifr/convective_or_high_wind), nas.impact_cascade (realized_delay_min + source_status + gate_passed + exposed_downstream hubs + air_cargo_exposure = downstream cross-validation, band none/contained/local_impact/propagating/network_cascade) - Sources: FAA NAS Status (nasstatus.faa.gov) + AviationWeather.gov (METAR/TAF) + Open-Meteo (weather fallback). Official/observable facts, screening-level, no judgment/prediction, NOT flight/booking/safety advice. Aviation weather is aviation-native (METAR ceiling/visibility) not generic precipitation. - Tracked: 22 US hub airports (EWR/LGA/JFK/SFO/ORD/BOS/PHL/ATL/DFW/DEN/CLT/IAH/MIA/SEA/LAX/DCA/MCO/LAS/PHX/MEM/SDF/ANC; MEM/SDF/ANC air-cargo hubs). NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 3 signals, no lifecycle or timeline. ## Land border wait triangle (mature real-time monitoring, not a death registry) Real-time US land-border crossing wait per port of entry from CBP BWT (Border Wait Times) for AI agents (cross-border trucking/3PL, JIT manufacturers — auto/electronics maquiladora, produce importers, customs brokers, supply-chain-risk bots). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): lane-capacity (commercial lanes/booths open below the maximum — deficit_ratio, lanes_open vs maximum_lanes, the structural driver, leading upstream) → crossing-wait (current CBP wait minutes — commercial + passenger + headline delay minutes, plus port status) → freight-delay-realized (commercial standard delay + FAST lane dual-corroboration confirming a FREIGHT-specific jam, with JIT-supply exposure). Confidence is read by how many stages are lit in causal order — booths reduced with no rising wait = a structural warning; all three lit = booths reduced then wait rising then freight-jammed confirmed by FAST plus commercial. This exact moment's CBP BWT wait minutes and port status an LLM cannot know live. This is LAND-border truck/car throughput identity — shipping = maritime line, nas = air line, tariff-policy = legal trade measures; agriculture/shipping/tariff-policy are the max-perimeter cascade cross-refs (produce-gateway spoilage / intermodal / trade-cost). Live-fetch, free no-key CBP BWT source. - Skill: /gauge/border-wait-region?entity= — $0.25, one port of entry: lane-capacity + crossing-wait + freight-delay-realized + sequential-confirmation - Signals (query via /gauge, /gauge/census): border-wait.lane-capacity (commercial lanes/booths open below maximum — deficit_ratio + lanes_open + maximum_lanes + stream + lanes_closed + port_closed = upstream leading, band full/minor/reduced/severe/shutdown), border-wait.crossing-wait (current CBP wait minutes + port status — current.wait + commercial_delay_min + passenger_delay_min + headline_delay_min + port_status + border + freight_crossing + produce_gateway = current, band open/light/moderate/heavy/severe/closed), border-wait.freight-delay-realized (commercial standard + FAST lane dual-corroboration confirming a FREIGHT jam — current.freight + commercial_delay_min + fast_delay_min + fast_lanes_closed + has_commercial + jit_supply_exposure + cascade_lines = downstream cross-validation, band clear/freight-strained/freight-jammed/freight-severe) - Sources: CBP Border Wait Times (bwt.cbp.gov/api/waittimes; free, no key). Official/observable facts, screening-level, no judgment/prediction, NOT travel/routing/customs advice. Land-border truck/car wait is distinct from shipping (maritime), nas (air) and tariff-policy (legal); agriculture (produce-gateway spoilage) / shipping (intermodal) / tariff-policy (trade-cost) are the freight/produce cascade cross-refs. NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 3 signals, no lifecycle or timeline. ## Drug shortage triangle (mature real-time monitoring, not a death registry) Real-time US drug-shortage burden per therapeutic category from the FDA/openFDA shortage list for AI agents (hospital / GPO pharmacy procurement, drug-supply-chain sourcing, PBMs and insurers, pharma competitive-intel, parametric supply-disruption). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): onset (drug-shortage.onset new-shortage-onset rate = upstream leading, the new-shortage-onset rate leads the burden peak) → burden (drug-shortage.burden current active-shortage burden per therapeutic category — active shortage count / unavailable count / unavailable share / criticality = current state) → entrenchment (drug-shortage.entrenchment downstream cross-validation — duration + discontinuation + no-availability confirming the shortage is entrenched, with procurement / filing / shipping / tariff cascade). Confidence is read by how many stages are lit in causal order (rising new onsets → high active burden → entrenched/attriting with discontinuation). This exact moment's openFDA shortage values an LLM cannot know live. Live-fetch collected on a real-time cadence, served from stored snapshots (data_age_hours/freshness labeled; the paid call reads stored SQL, it does not trigger a fresh upstream fetch), free openFDA source (optional key for rate limits only). Cross-line cascade: cascade_lines (filing manufacturer-distress, shipping + tariff API-import) plus med_device_xref (drug-device combo, cross-ref only) as the max-perimeter cascade. HONESTY CAVEAT: records report FDA-listed status, availability and dates verbatim; this is screening intel (which category is under supply stress, how many drugs, how long, resolving vs discontinuing), NOT medical, clinical or treatment advice. - Skill: /gauge/drug-shortage-region?entity= — $0.25, one therapeutic category: onset + burden + entrenchment + sequential-confirmation - Signals (query via /gauge, /gauge/census): drug-shortage.burden (current.burden + active_shortage_count + unavailable_count + unavailable_share + therapeutic_category + criticality = current, bands none/low/elevated/high/severe), drug-shortage.onset (current.onset + new_onset_count + current_total + onset_share = upstream leading new-shortage-onset rate, bands quiet/stirring/rising/surging/spiking), drug-shortage.entrenchment (current.entrenchment + max_duration_days + discontinuing_share + unavailable_share + active_count + discontinuing_count + procurement_stress + manufacturer_distress_candidate + api_import_exposure + cascade_lines + med_device_xref = downstream cross-validation, bands transient/sticky/entrenched/attriting/critical-loss) - Sources: openFDA Drug Shortages (api.fda.gov/drug/shortages; free live source, optional key for rate limits only) + FDA Drug Shortages web database + ASHP shortage list (corroborants). Official openFDA provenance, screening-level, reports FDA-listed status/availability/dates verbatim, no judgment/prediction, NOT medical/clinical/treatment advice. Boundary: med-device (the X line) is FDA device recalls, this is drug supply availability; the manufacturer-distress filing and the shipping + tariff API-import are cascade cross-refs, and the drug-device combo med_device_xref is cross-ref only. - Tracked: 18 therapeutic categories (anesthesia, oncology, psychiatry-CNS, analgesia-opioid, anti-infective, cardiovascular, neurology, gastroenterology, endocrine, hematology, pulmonary-allergy, ophthalmology, TPN-nutrition, renal, medical-imaging, antiviral, pediatric, dermatology). NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 3 signals, no lifecycle or timeline. ## Per-drug shortage + alternatives triangle (mature real-time monitoring, not a death registry) Real-time PER-DRUG US FDA shortage detail with NDC / manufacturer granularity plus a therapeutic-alternative (ATC-class) shortage cross-validation, per drug, from openFDA + RxNav/RxClass for AI agents (hospital / GPO procurement that buys NDCs not categories, short-term drug-supply researchers). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): supply-narrowing (drug-detail.supply-narrowing available_manufacturers / total_manufacturers_listed = upstream leading, fewer available manufacturers is a fragility that leads the shortage) → status (drug-detail.status current per-drug FDA shortage state + affected_ndc_count / manufacturer_count / unavailable_share / presentations / drug_class = current, the per-NDC granular state) → alternatives-status (drug-detail.alternatives-status downstream cross-validation — are the same-ATC-class therapeutic alternatives atc_class / alternatives_list ALSO in FDA shortage = fallback exhausted, with alternatives_also_short / alternatives_short_count / self_in_shortage / clinical_supply_exposure and cascade to hospital-procurement / drug-shortage-category / manufacturer filing) → alternative-supply (drug-detail.alternative-supply resolution leg — of those ATC-class alternatives, WHICH specific manufacturers/NDCs are still marketed and buyable per openFDA /drug/ndc: viable_alternatives / viable_count / procurement_exposure, so procurement knows not just that a fallback exists but that it actually ships). Confidence is read by how many stages are lit in causal order (available manufacturers narrowing → the drug listed in shortage → the ATC-class alternatives also short = fallback exhausted). This exact moment's openFDA per-drug shortage and RxNav alternatives an LLM cannot know live. Live-fetch collected on a real-time cadence, served from stored snapshots (data_age_hours/freshness labeled; the paid call reads stored SQL, it does not trigger a fresh upstream fetch), free openFDA + NIH/NLM RxNav sources (no key). This is the PER-DRUG + SUBSTITUTION view, finer than the category-level drug-shortage line. HONESTY CAVEAT: records report the FDA shortage list and same-ATC-class drugs verbatim; this is screening intel, NOT medical, clinical, substitution or treatment advice. - Skill: /gauge/drug-detail-region?entity= — $0.25, one drug: supply-narrowing + status + alternatives-status + alternative-supply (WHICH specific manufacturers/NDCs of the alternatives still ship) + sequential-confirmation - Signals (query via /gauge, /gauge/census): drug-detail.status (current.status + current_records + affected_ndc_count + manufacturer_count + unavailable_share + presentations + drug_class = current per-drug FDA shortage state, bands available/limited/in-shortage/severe-shortage/discontinuing), drug-detail.supply-narrowing (current.supply + available_manufacturers + total_manufacturers_listed = upstream leading fragility, bands broad/adequate/narrow/sole-source/none), drug-detail.alternatives-status (current.alternatives + atc_class + alternatives_list + alternatives_also_short + alternatives_short_count + alternatives_total + self_in_shortage + clinical_supply_exposure + cascade_lines = downstream cross-validation of whether the ATC-class alternatives are also short, bands alternatives-ok/some-alts-short/most-alts-short/no-safe-alternative/no-alternatives), drug-detail.alternative-supply (current.fallback + atc_class + self_in_shortage + viable_alternatives[{generic, labeler_count, ndc_count, still_shipping}] + viable_count + alternatives_total + procurement_exposure + cascade_lines = resolution leg — of the ATC-class alternatives, WHICH specific manufacturers/NDCs are still marketed and buyable per openFDA /drug/ndc, bands multi-source-available/single-source-available/thin-supply/no-shipping-alternative/no-alternatives) - Sources: openFDA Drug Shortages (api.fda.gov/drug/shortages, per-drug; free no-key live) + RxNav / RxClass ATC therapeutic classification (NIH/NLM; free no-key) + FDA Drug Shortages database (corroborant). Official openFDA + NIH/NLM provenance, screening-level, reports the FDA shortage list and same-ATC-class drugs verbatim, no judgment/prediction, NOT medical/clinical/substitution/treatment advice. Boundary: the drug-shortage sibling line is the therapeutic-category level, THIS line is per-drug/NDC detail plus alternatives; hospital-procurement, the drug-shortage category line and the manufacturer filing are the cascade cross-refs. - Tracked: 24 drugs (carboplatin, cisplatin, oxaliplatin, methotrexate, fluorouracil, epinephrine, norepinephrine, phenylephrine, atropine, amoxicillin, piperacillin-tazobactam, cefazolin, vancomycin, heparin, dextrose, sodium-chloride, sodium-bicarbonate, lidocaine, ketamine, morphine, hydromorphone, fentanyl, magnesium-sulfate, dexamethasone). NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 4 signals (supply-narrowing → status → alternatives-status → alternative-supply), no lifecycle or timeline. ## Earthquake seismic-activity triangle (mature real-time monitoring, not a death registry) Real-time global earthquake activity & impact per seismic zone from USGS for AI agents (supply-chain/logistics planners, parametric insurers, infrastructure operators, humanitarian logistics). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): swarm-rate (small-event rate ratio vs baseline = upstream leading, a DESCRIPTIVE elevated small-event-rate signal) → current-activity (current USGS state — max magnitude / event count / max significance / shallowest depth / strongest place / tectonic setting / tsunami-prone) → realized-impact (ShakeMap MMI + PAGER alert & population exposure + tsunami flag with NOAA/PTWC cross-reference + felt reports + aftershock risk = downstream cross-validation). Confidence is read by how many stages are lit in causal order. This exact moment's USGS values an LLM cannot know live. Live-fetch collected on a real-time cadence, served from stored snapshots (data_age_hours/freshness labeled; the paid call reads stored SQL, it does not trigger a fresh upstream fetch), free no-key USGS sources. HONESTY CAVEAT: earthquakes are NOT reliably predicted; the swarm-rate leg is a descriptive elevated small-event-rate signal, NOT a forecast of a larger quake; the tsunami (NOAA/PTWC) and PAGER population exposure are max-perimeter cascade cross-refs. - Skill: /gauge/earthquake-region?entity= — $0.25, one seismic region: swarm-rate + current-activity + realized-impact + sequential-confirmation - Signals (query via /gauge, /gauge/census): earthquake.current_activity (current.max_magnitude + event_count + max_significance + shallowest_depth_km + strongest_place + tectonic + tsunami_prone = current, bands quiet/minor/notable/strong/major), earthquake.swarm_rate (current.rate_ratio + current_count + baseline_expected = upstream leading descriptive precursor, NOT a prediction, bands baseline/slightly_elevated/elevated/swarm), earthquake.impact_realized (current.impact + event_id + event_mag + pager_alert + max_mmi + tsunami_flag + felt_reports + tsunami_cross_ref + pager_population_exposure + adjacent_aftershock_risk = downstream cross-validation, bands none/felt/moderate/significant/severe) - Sources: USGS FDSN event query (earthquake.usgs.gov/fdsnws/event) + USGS Earthquake GeoJSON summary feed + USGS ShakeMap/PAGER/DYFI (ShakeMap MMI, PAGER alert & population exposure, DYFI felt reports) + NOAA/PTWC as tsunami cross-reference pointer. Official USGS provenance, screening-level, no judgment/prediction, NOT emergency/evacuation/safety advice. Earthquakes are not reliably predicted; the swarm-rate leg is descriptive (elevated small-event rate), NOT a prediction of a larger quake; the tsunami and PAGER population exposure are cascade cross-refs. - Tracked: 18 global seismic zones (California San Andreas, Cascadia PNW, Alaska-Aleutians, Japan, Sumatra-Indonesia, Philippines, Chile, Mexico, Central America, Turkey-Anatolia, Greece-Italy-Med, Iran, Nepal-Himalaya, New Zealand, Iceland, PNG-Solomon, Taiwan, Caribbean). NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 3 signals, no lifecycle or timeline. ## Grid reliability / outage-risk triangle (mature real-time monitoring, not a death registry) Real-time US grid reliability and operational strain per Balancing Authority from EIA-930 for AI agents (data-center / AI-compute site-selection, energy-intensive industrials, demand-response and energy-trading, parametric-outage insurers). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): forecast-stress (upstream leading — EIA-930 day-ahead demand forecast DF approaching the recent actual peak, about a 1-day lead) → stress (current EIA-930 operational strain — demand vs recent-peak headroom, net generation, interchange, import share) → realized-strain (downstream cross-validation — forecast error + import dependence + demand-beat-forecast, confirming the grid RAN TIGHT, with ai-compute data-center power/curtailment and industrial cascade). Confidence is read by how many stages are lit in causal order (day-ahead forecast tight → current headroom thin → realized strain confirms). Live-fetch collected on a real-time cadence, served from stored snapshots (data_age_hours/freshness labeled; the paid call reads stored SQL, it does not trigger a fresh upstream fetch) EIA-930, KEY-GATED — dormant and empty until a free EIA_API_KEY is set (like wildfire needs a FIRMS key); this exact hour's EIA-930 demand/forecast/interchange values an LLM cannot know live. HONESTY CAVEAT: EIA-930 has NO outage feed — the realized-strain leg confirms operational strain (the grid ran tight), it is NOT a literal power-outage claim. Cross-line cascade to ai-compute (data-center power / curtailment exposure) and industrial, with a weather-driver cross-ref (severe-convection / wildfire). Boundary: the existing electricity line is an ERA5 degree-days demand proxy plus prices; THIS line is actual EIA-930 grid operations. Screening-level, official EIA provenance, record_hash verifiable (canonical sha256, recomputable offline), NOT dispatch / operational / safety advice. - Skill: /gauge/grid-reliability-region?entity= — $0.25, one Balancing Authority: forecast-stress + stress + realized-strain + sequential-confirmation - Signals (query via /gauge, /gauge/census): grid-reliability.stress (current.stress + demand_mw + recent_peak_mw + headroom_pct + net_generation_mw + total_interchange_mw + import_share + region = current, bands normal/elevated/tight/critical/emergency), grid-reliability.forecast-stress (current.forecast_band + forecast_tightness_ratio + df_peak_mw + recent_actual_peak_mw = upstream leading day-ahead demand forecast approaching peak, bands slack/watch/stressed/peak-risk/over-peak), grid-reliability.realized-strain (current.strain + forecast_error + import_share + over_forecast + demand_mw + forecast_mw + datacenter_exposure + industrial_curtailment_exposure + cascade_lines + weather_driver_xref = downstream cross-validation confirming the grid ran tight NOT a literal outage, bands confirmed-slack/nominal/strained/ran-tight/severe-strain) - Sources: EIA-930 Hourly Grid Monitor (EIA API v2 electricity/rto/region-data — hourly demand D, day-ahead demand forecast DF, net generation NG, total interchange TI by Balancing Authority; free EIA_API_KEY, key-gated) + DOE OE-417 Electric Emergency Incident and Disturbance Report (non-real-time outage cross-reference, mention-only). Official EIA provenance, screening-level, no judgment/prediction, NOT dispatch/operational/safety advice. EIA-930 has no outage feed; realized-strain confirms operational strain (ran tight) not a literal outage; boundary electricity = ERA5 degree-days demand proxy + prices, this = actual EIA-930 grid operations. - Tracked: 18 US Balancing Authorities (CAISO, ERCOT, MISO, PJM, SPP, NYISO, ISO-NE, BPA, PacifiCorp East/West, FPL, Southern Co, TVA, Duke, PSCo, AZPS, AECI, PNM). KEY-GATED — needs a free EIA_API_KEY, dormant/empty until set. NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 3 signals, no lifecycle or timeline. ## Severe convection / hail triangle (mature real-time monitoring, not a death registry) Real-time US severe convective storm hazard (hail / damaging wind / tornado) per region from NWS/SPC for AI agents (property / crop / parametric-hail insurers, agriculture and logistics, aviation and fleet, utilities). A SEQUENTIAL-CONFIRMATION chain (NOT a multiplied score): forecast-risk-env (upstream leading — SPC day-1 categorical convective outlook + CAPE/CIN instability + wind gust/shear, hours-to-1-day lead) → active-warning-state (current NWS severe-thunderstorm and tornado warnings + watches + PDS emergency) → realized-severe-reports (downstream cross-validation — SPC realized max hail size / max wind / tornado count + hail/wind reports + crop, aviation and power exposure + cross-line cascade). Confidence is read by how many stages are lit in causal order (elevated SPC/CAPE risk leading → NWS warning issued → realized severe reports confirm). Live-fetch collected on a real-time cadence, served from stored snapshots (data_age_hours/freshness labeled; the paid call reads stored SQL, it does not trigger a fresh upstream fetch), free no-key NWS/SPC sources; this exact moment's NWS and SPC values an LLM cannot know live. Cross-line cascade: agriculture crop-hail, nas aviation-hail, electricity wind-tornado-outage = max-perimeter cascade. Boundary: precipitation is rain totals; wildfire fire-weather is hot and dry (the opposite); this is severe-storm hazard (hail / wind / tornado). Screening-level, official NWS/SPC provenance, record_hash verifiable (canonical sha256, recomputable offline), NOT emergency/shelter/safety advice. - Skill: /gauge/severe-convection-region?entity= — $0.25, one convective region: forecast-risk-env + active-warning-state + realized-severe-reports + sequential-confirmation - Signals (query via /gauge, /gauge/census): severe-convection.active-warning-state (current.hazard_tier + active_warnings + active_watches + tornado_warnings + svr_warnings + count_band + regime = current, bands none/watch/svr-warning/tornado-warning/pds-emergency), severe-convection.forecast-risk-env (current.risk_band + spc_category + spc_dn + cape_jkg + cin_jkg + wind_gust_mph + source_used = upstream leading, bands tstm/marginal/slight/enhanced/moderate/high), severe-convection.realized-severe-reports (current.impact + max_hail_in + max_wind_mph + tornado_count + hail_reports + wind_reports + crop_exposure + aviation_exposure + power_exposure + cascade_lines = downstream cross-validation, bands none/marginal-severe/significant-hail/significant-wind/tornado-confirmed/violent) - Skill: /gauge/hurricane-region?entity= — $0.25, one NHC basin: formation-outlook + active-storms + impact-cascade + season-timeline + sequential-confirmation. 🚨 DESCRIPTIVE ONLY — official NHC/NOAA tropical-cyclone data, NOT safety/evacuation/emergency advice; for warnings & evacuation consult NHC (nhc.noaa.gov), NWS and local emergency management. - Skill: /gauge/drought-region?entity= — $0.25, one US climate region: onset-pressure (evaporative demand / flash-drought, Open-Meteo ET0, leads USDM by weeks) + status (OFFICIAL US Drought Monitor D0-D4 + %area) + impact-cascade (realized hydro/socioeconomic gate authorizing edges → agriculture/electricity-hydropower/shipping-river-barge/wildfire-fuel/hydrology/energy/filing) + season-timeline + sequential-confirmation. ★DISTINCT from agriculture.drought (ERA5 soil-moisture proxy) — this is the official integrative classification + socioeconomic cascade. Official USDM/NOAA/USGS, honest-empty when no drought. NOT water-management/irrigation/investment advice. - Signals (query via /gauge, /gauge/census): drought.status (current.band + usdm_valid_date + pct_area_D0plus..D4 + pct_area_D2plus = official USDM, bands none/D0-abnormally-dry/D1-moderate/D2-severe/D3-extreme/D4-exceptional; DMData API), drought.onset-pressure (current.band + et0_30d_mm + precip_30d_mm + demand_deficit_mm = upstream leading evaporative demand, bands nominal/ED0..ED3; Open-Meteo ET0, official EDDI/NOAA CPC reference), drought.impact-cascade (current.band + status_gate + pct_area_D2plus + streamflow_cfs + cascade_targets = downstream realized gate + edges, bands no-impact/emerging/significant/severe/extreme; USGS+USDM, honest-empty when no drought), drought.industry-timeline (current.season_position + latest_peak_pct_d2plus + by_year = multi-year history, bands below-normal/normal/above-normal/severe-widespread; USDM history+NOAA) - Signals (query via /gauge, /gauge/census): hurricane.active-storms (current.band + active_storm_count + strongest + storms[saffir_cat_band/max_wind_kt/min_pressure_mb/lat/lon/in_gulf] = current, bands none/depression/tropical-storm/cat1-2/major; NHC CurrentStorms.json), hurricane.formation-outlook (current.band + prob_2day_pct_max + prob_7day_pct_max + disturbances = upstream leading days ahead, bands none/low/medium/high; NHC GTWO, quoted official value), hurricane.impact-cascade (current.gate + realization + warning_count + watch_count + gulf_storm_present + cascade_targets = downstream cross-validation, gate none/watch/warning; official surge/wind watches-warnings authorize geo cascade edges → energy/electricity/shipping/nas/agriculture/hydrology/precipitation/filing, honest-empty offshore/off-season), hurricane.industry-timeline (current.season_position + latest_counts + normals + by_season = season history, bands below-normal/normal/above-normal/hyperactive; NOAA/NHC HURDAT2) - Sources: NWS API (api.weather.gov active alerts — severe-thunderstorm/tornado warnings + watches) + SPC Storm Prediction Center (day-1 categorical outlook + storm reports for realized hail/wind/tornado) + Open-Meteo CAPE (convective instability). Official NWS/SPC provenance, screening-level, no judgment/prediction, NOT emergency/shelter/safety advice. Boundary: precipitation = rain totals; wildfire fire-weather = hot/dry (opposite); this = severe-storm hazard. - Tracked: 16 US severe-convective regions (Tornado Alley core, TX Panhandle hail, Central Kansas, High Plains hail alley, Front Range CO, Northern Plains, Upper Midwest derecho, Corn Belt IL, Ohio Valley, Tennessee Valley, Dixie Alley core, ArkLaTex, Southern Plains TX, TX Hill Country, Southeast Atlantic, Florida Peninsula). NOTE: causal-chain signals use sequential-confirmation (are the stages lit in order), never a multiplied score. Mature line — 3 signals, no lifecycle or timeline. ## DeFi-lending triangle — lending-protocol solvency & exploit forensics (emerging industry · DeFi cascade HUB) DeFi lending / money-market death forensics for AI agents (lending-market depositors/LPs, yield-allocating agents auto-routing into high-APY markets, protocols composing on top, risk agents). Protocols that go insolvent (bad-debt socialization), get exploited/hacked, or freeze/deprecate markets. Three legs per protocol + cross-validation. Complete cycle (emerging industry). THIS LINE IS THE DEFI CASCADE HUB: it is the SINK for the lrt (LRT depeg → bad debt) and rwa (RWA impair → bad debt) collateral cascades, and the SOURCE feeding the stablecoin line (an insolvent CDP issuer's DAI/GHO-class coin de-peg risk). Moat: the pre-death bad-debt/utilization/oracle time-series is erased the moment a protocol freezes or is drained — un-reconstructable after collapse. Boundary: crypto over-collateralized money-markets (distinct from rwa off-chain private-credit pools; distinct from stablecoin peg — the CDP coin is only a cascade edge). NOT investment advice. - Skill: /gauge/defilending-region?entity= — $0.25, one protocol: lifecycle + bad-debt stress + contagion cascade + cross-check - Signals (query via /gauge, /gauge/census): defi-lending.lifecycle (operating → at_risk → exploited/insolvent/deprecated → dead; by protocol-type/chain), defi-lending.bad-debt-stress (un-liquidated underwater positions + utilization + oracle deviation = upstream leading; where LRT/RWA collateral impairments first surface as bad debt), defi-lending.contagion-cascade (exploit/insolvency kill + outbound cascade to stablecoin CDP coins + depositor flight/cross-protocol; gated by on-chain kill event = downstream cross-validation), defi-lending.industry-timeline (complete cycle by death-type/chain; 2022–2024 exploit wave) - Sources: on-chain TVL/bad-debt/utilization (DefiLlama + on-chain reserve/account state) + oracle-vs-DEX deviation + governance risk-actions + exploit tx + Rekt/SlowMist post-mortems. Observable/on-chain facts, no judgment/prediction, no investment advice. - Tracked: ~20 protocols (Aave, Compound, Morpho, Euler v2, Sky/Maker, Spark, Fluid, Silo, Venus, Kamino, Save/Solend, Inverse + dead Euler v1/Cream/Rari-Fuse/Radiant/UwU-Lend/bZx/Hundred/Iron-Bank); operating/at_risk/ended/dead by chain (Ethereum/Arbitrum/BSC/Solana). ## AI-dataset triangle — training-dataset takedown & provenance forensics (emerging industry) AI training-dataset death forensics for AI agents (model builders/fine-tuners whose corpus was pulled → reproducibility + legal exposure, AI-compliance/audit agents doing EU AI Act training-data transparency, model-provenance/SBOM-for-AI agents, legal DD agents). Training datasets/corpora pulled/gated/deleted via copyright/DMCA (Books3), CSAM/safety takedown (LAION-5B), privacy/facial withdrawal (MS-Celeb-1M/MegaFace/DukeMTMC/80M-Tiny-Images), court order, license revoke, or ethics retraction. Three legs per dataset + cross-validation. Complete cycle (emerging industry). KILLER PROPERTY: a deleted dataset is DEFINITIONALLY un-reconstructable — once pulled, the exact corpus a model trained on is gone; the pre-death hash-anchored snapshot is the only surviving proof of what the exposed models actually ingested. MAX cross-line cascade: a dead dataset → the models/tools/promises trained on it (ai-deps/ai-tools/promises). Regulatory tailwind: EU AI Act training-data provenance. NOT legal advice. - Skill: /gauge/aidataset-region?entity= — $0.25, one dataset: lifecycle + legal-distress + model-lineage-exposure + cross-check - Signals (query via /gauge, /gauge/census): ai-dataset.lifecycle (available → gated/disputed → removed/deleted; by modality/host), ai-dataset.legal-distress (DMCA/Lumen notice, copyright/privacy lawsuit, regulator/DPA letter, ethics retraction = upstream leading, precedes takedown by weeks-to-months), ai-dataset.model-lineage-exposure (which models trained on the dead dataset → provenance/legal exposure → cascade into ai-deps/ai-tools/promises; gated by repo-404 + Wayback removal proof = downstream cross-validation), ai-dataset.industry-timeline (complete cycle by takedown-reason/modality; privacy-facial wave 2019–20 vs copyright/CSAM wave 2023) - Sources: HuggingFace/GitHub/Kaggle availability + license + Lumen DMCA notices + court dockets + Wayback Machine + model-card training-data disclosures. Observable/legal facts, no judgment/prediction, no legal advice. - Tracked: ~19 datasets (LAION-5B [CSAM takedown]→Re-LAION, Books3 [DMCA], MS-Celeb-1M/MegaFace/DukeMTMC/Brainwash/80M-Tiny-Images [privacy/ethics withdrawals] + at-risk The-Pile/ImageNet/LAION-400M/RedPajama/The-Stack/Clearview + operating Common-Crawl/C4/Common-Voice/OpenWebText); operating/at_risk/ended/dead by modality. ## Med-device recall triangle — medical-device recall / withdrawal & adverse-event forensics (emerging industry) Medical-device recall/withdrawal death forensics for AI agents (hospital procurement/med-supply-chain risk agents, device due-diligence agents, payers/insurers, product-liability DD). Device product-lines that get recalled, withdrawn, or have their authorization revoked via FDA/EUDAMED — Class I recalls (Philips DreamStation CPAP foam), device withdrawals (Medtronic HeartWare HVAD), and implant recalls escalating into MDL litigation (Exactech knee/hip, DePuy ASR hip, Allergan BIOCELL, Bayer Essure). Three legs per product-line + cross-validation. Complete cycle (emerging industry). LEADING PROPERTY: the MAUDE adverse-event surge leads the recall by months — reports pile up before the FDA classifies. Boundary: the med-device recall is the IDENTITY of this line; the manufacturer 8-K is only a cascade edge INTO the filing line, not this line identity. NOT medical or legal advice. - Skill: /gauge/meddevice-region?entity= — $0.25, one product-line: lifecycle + adverse-event surge + recall-cascade + cross-check - Signals (query via /gauge, /gauge/census): med-device.lifecycle (cleared → recall → withdrawn/revoked; 510(k)/PMA/EUDAMED, by device-type/class/manufacturer), med-device.adverse-event (MAUDE current.surge + severity_score, band quiet/elevated/surging/pre-recall-critical = upstream leading, leads recall by months), med-device.recall-cascade (FDA Enforcement recall classification GATE, current.cascade + recall_class/corporate/litigation/procurement, band unconfirmed/fda-classified/corporate-cascade/litigation-cascade/full-cascade → manufacturer 8-K filing + MDL litigation + GPO procurement withdrawal = downstream cross-validation), med-device.industry-timeline (complete cycle by death-type/device-class) - Sources: openFDA device endpoints (recall + enforcement classification + event MAUDE + 510(k) + PMA) + EU EUDAMED (authorization + field safety notices) + manufacturer 8-K (cascade edge) + MDL dockets + GPO procurement. Official/observable facts, no judgment/prediction, no medical/legal advice. - Tracked: ~20 device product-lines (Philips DreamStation CPAP / Respironics Trilogy Evo, Medtronic HeartWare HVAD / MiniMed 600 / Micra AV / Hugo RAS, Exactech Optetrak knee / Connexion GXL hip, DePuy ASR hip, Smith+Nephew Birmingham hip, Allergan BIOCELL implant, Bayer Essure, Bard/Cook IVC filters, Abbott Nanostim/St-Jude ICD/FreeStyle Libre 3, Boston Scientific Lotus Edge TAVR, Dexcom G7 CGM, Intuitive da Vinci 5); cleared/recall/withdrawn/revoked by device class. ## Tariff/customs policy triangle — tariff / customs trade-measure legal-force & trade-flow forensics (emerging industry) Tariff/customs trade-measure legal-force death forensics for AI agents (procurement / importer-risk / trade-compliance due-diligence agents, customs brokers, supply-chain risk agents). Trade measures that get imposed then expire, are rescinded, struck down in court, or superseded — Section 301/232/201 tariffs, AD/CVD orders, IEEPA tariffs, de-minimis changes — via USTR / Federal Register / USITC / WTO / CIT. Three legs per measure + cross-validation. Complete cycle (emerging industry). LEADING PROPERTY: the review-pipeline (sunset review / CIT-WTO court challenge / exclusion expiry) leads the measure death by months — official dated events precede the legal-force change. Boundary: the trade/customs measure is the IDENTITY of this line; E climate-policy covers carbon/climate instruments, and the importer 8-K, shipping reroute and agriculture retaliation are cascade edges INTO other lines, not this line identity. NOT legal, investment or customs advice. - Skill: /gauge/tariff-region?entity= — $0.25, one trade-measure: legal-force lifecycle + review-pipeline leading + trade-flow cascade + cross-check - Signals (query via /gauge, /gauge/census): tariff-policy.lifecycle (imposed → expired/rescinded/struck-down/superseded; measure_type/jurisdiction/hs_scope/legal_basis, Section 301/232/201, AD/CVD, IEEPA, de-minimis), tariff-policy.review-pipeline (current.stage + event, band none/review_scheduled/review_initiated/challenge_filed/determination_pending/terminal_action_imminent = upstream leading, sunset review / CIT-WTO challenge / exclusion expiry, official dated events lead death), tariff-policy.tradeflow-cascade (Federal Register final-action GATE, current.cascade + fr_confirmed/tradeflow/exposed_lines, band unconfirmed/fr-confirmed/tradeflow-shifted/multi-line-cascade → USITC DataWeb import-flow shift → exposed importers/manufacturers 8-K filing + shipping origin reroute + agriculture retaliation = downstream cross-validation), tariff-policy.industry-timeline (complete cycle by death-type/measure-type) - Sources: USTR + US Federal Register API + USITC (DataWeb import trade-flow + AD/CVD sunset reviews & orders) + WTO (dispute settlement) + US CBP (customs rulings/enforcement) + EU TARIC (integrated tariff). Official/observable facts, no judgment/prediction, no legal/investment/customs advice. - Tracked: ~22 trade-measures (Section 301 China lists 1/3/4A + EV modifications + exclusions, Section 232 steel/aluminum/EU-TRQ, Section 201 solar/washers/steel-2002, GSP program, NAFTA, Airbus-Boeing tariffs, AD/CVD solar-cells-China/shrimp/honey-China/oil-country-tubular, IEEPA reciprocal-2025/fentanyl-tariffs-2025, de-minimis change, EU CBAM customs); imposed/review/expired/rescinded/struck-down/superseded by measure type. ## Biodiversity credit triangle — biodiversity / nature-credit integrity & market-existence forensics (nascent industry, honesty posture) Biodiversity / nature-credit integrity and market-existence screening for AI agents (procurement / ESG / corporate nature-pledge / due-diligence agents). Nature/biodiversity credit projects, schemes and methodologies that get registered and issue, drift into at_risk, then can be suspended, delisted, have the methodology revoked, see the scheme collapse, or go demand-dead — UK BNG statutory units, Wallacea/Plan Vivo/Verra/Terrasos/Cercarbono methodologies, ecoAustralia/Savimbo/CreditNature/Ekos credits. Three legs per credit + cross-validation. Complete cycle (nascent industry). LEADING PROPERTY: the MRV-integrity drift (methodology-under-review / additionality challenge / rating downgrade / greenwash expose — the multidimensional habitat/species/condition MRV gap) is the upstream leading signal; an at_risk credit is corroborated as integrity-fragile when MRV-drift is elevated, and a terminal death requires the official standard-body gate. HONESTY POSTURE: this is a under-5-year-old market that is mostly operating or at_risk with ZERO confirmed permanent deaths, and the buyer-holder cascade schema is wired but honestly empty (holders and exposed_lines empty) — no fabricated fund exposure. Boundary: the nature/biodiversity credit instrument is the IDENTITY of this line; carbon(A) covers single-tonne CO2e credits, esg-fund(Q) covers the fund vehicle, and co-located carbon projects / fund holdings / corporate nature-pledges are cascade edges INTO other lines, not this line identity. Screening-level, NOT investment or environmental advice. - Skill: /gauge/biodiversity-region?entity= — $0.25, one biodiversity credit: integrity & market-existence lifecycle + MRV-integrity drift leading + buyer-holder cascade + cross-check - Signals (query via /gauge, /gauge/census): biodiversity-credit.lifecycle (registered/issuing → at_risk → suspended/delisted/methodology-revoked/scheme-collapsed/demand-dead; credit_type/jurisdiction/standard_body/metric_basis), biodiversity-credit.mrv-integrity-drift (current.drift + severity_score, band sound/watch/challenged/under-review/impaired = upstream leading, methodology review / additionality / rating / greenwash, multidimensional MRV), biodiversity-credit.buyer-holder-cascade (official standard-body terminal instrument OR documented demand-dead registry flatline GATE, current.cascade + official/holders/exposed_lines, band none/demand-thin/demand-dead/official-terminal/holder-exposed → esg-fund funds holding it / promises corporate nature-pledges / carbon co-located projects = downstream cross-validation, wired but honestly empty), biodiversity-credit.industry-timeline (complete cycle by death-type) - Sources: standard-body registries (Verra / Plan Vivo / Terrasos / Cercarbono) + UK BNG register (Natural England) + Biodiversity Credit Alliance (BCA) + TNFD. Official/observable facts, no judgment/prediction, no investment/environmental advice. - Tracked: ~18 biodiversity/nature credits & schemes (UK BNG statutory units + register, Wallacea methodology, Plan Vivo PV Nature, Verra nature framework, Terrasos habitat banks, ecoAustralia credits, Savimbo credits, CreditNature NARIA, Ekos credits, BCA governance, Gold Standard biodiversity, South Pole nature, Cercarbono biodiversity, Operation Wallacea legacy, BNG broker marketplaces, nature-credit index pilots, self-issued voluntary); registered/issuing/at_risk/suspended/delisted/methodology-revoked/scheme-collapsed/demand-dead. Nascent market, mostly operating/at_risk, zero confirmed permanent deaths, cascade wired but honestly empty. ## Death registries (vanished-data forensics) Verifiable forensic records of things that died: stablecoins (depeg deaths + x402 service trust scores). Free preview per line; paid per-record + census. (AI-tool / carbon / promise / dependency / policy forensics are now served via the unified GAUGE /gauge/*-region lines above.) - Free previews: /stablecoin-deaths/preview · /trust/preview ## Verify (zero-trust) — provenance is scoped, not the headline Every record ships record_hash (canonical sha256 over a versioned canonical core, hash_core_version=3) + official source_ref/timestamp. Recompute it yourself offline from the delivered payload, or cross-check for free: POST /verify (you hold the record) · GET /gauge/verify?hash=<record_hash> (you hold only the hash). Delivered payloads disclose their own recomputability in record_hash_verifiable. RETIRED — do not build against these: Merkle inclusion proofs and OpenTimestamps / Bitcoin anchoring were discontinued on 2026-07-23 (never stamped in production). /merkle/proof returns 410; /merkle/spec is archival only; ots_status is a compatibility field that stays 'pending' forever. /proof/:record_hash still returns integrity + official-timestamp provenance, and reports 'undetermined' (503) rather than 'not found' when a lookup cannot complete. ## Keywords official government data, US federal agency data, government records API, regulatory filing data, official statistics, flood risk, river water level, gage height, streamflow, flood stage, USGS, NOAA, NWS, EPA, SEC EDGAR, openFDA, EIA, BLS, FAA, US CBP, USTR, USITC, Federal Register, IMF PortWatch, air quality, US AQI, PM2.5, precipitation, drought, verifiable data, cryptographic proof, cryptographically signed citation, Ed25519, tamper-evident, auditable offline, provenance, record hash, canonical sha256, cross-source corroboration, leading indicator, x402, pay-per-call, USDC, Base, Solana, AI agent, no API key, no data no charge