Frequently asked
What is Nostradamus Intellect?
A predictive-intelligence observatory and public instrument of record. Each forecast is a written judgment — decomposed on its card, anchored to data and to earlier seals — carrying an explicit probability, a resolution date and a falsifiability condition, sealed before the outcome and graded in public. The engine — the Core — reads twelve statistical series and eight live wires and records how far the world has moved since each number was sealed. It tracks the numbers; it does not write them.
How does the Core actually work?
Four stages. Ingest: twenty tracked feeds — twelve absorbed statistical series (FRED, World Bank, SIPRI, the GPR index, Our World in Data and others) and eight live wires — are normalized. Pattern-match: the present is scored against indexed historical episodes on five components — debt load, great-power rivalry, internal polarization, technology shock, monetary strain — read through six macro-historical frameworks. Simulate: probabilistic scenario branches with widening confidence bands. Calibrate: probabilities are sealed, timestamped, and scored in public. Every weight is published in the Model Card.
Are these real Nostradamus predictions?
No — and that is the point. Nothing on this site is presented as a decoded quatrain. The original verses were unfalsifiable by design; that was the sixteenth century’s technology of authority, and it is exactly what this platform was built to answer. We keep the ambition of seeing far and add the accountability the original never faced: every projection carries a probability, a named adjudicator, a resolution date, and the observable condition that would prove it wrong. The Library quotes genuine quatrains, cited by Century and quatrain number, as heritage — and the Counter-Canon shelf sits beside them, cataloguing seven historical prophecy texts by their failure mode (hindsight retrofit, prophecy-after-the-event, motivated reasoning, outright forgery — Charles Hindley confessed the Mother Shipton fabrication in Notes & Queries in April 1873) and pairing each one with the mechanism here that neutralises it. We are the honest opposite of prophecy, built in its shape.
Which historical era does today most resemble?
The Historical Resonance Index scores nine eras against July 2026. Strongest matches: 1968–1974 (83/100), then 1910–1914 and 1873–1896 tied at 82 — the first a monetary-and-polarization rhyme, the second a pre-war rivalry rhyme, the third an economic installation-period rhyme. They are followed by 1929–1933 (79), 1848–1849 (79), 1937–1939 (78) and 1815–1819 (77, the only era carrying today’s debt load). The scores decompose into five published components with public weights — resonance identifies the fork, not the branch.
How accurate are the predictions?
Unknown — and we refuse to pretend otherwise. Sixteen predictions are sealed in the Calibration Ledger across two batches — ten at launch on July 3, 2026, and six near-term tests on July 11, 2026 that resolve by end-2027 — alongside twenty-eight live projections in the observatory. All are PENDING. The first sealed entries resolve as early as January–February 2027 (the 2026 warmest-year ranking), and the first formal public Brier review arrives in July 2027. Until then, the only honest track record is the seal itself.
What is a calibrated prediction?
A forecaster is calibrated when the things they call 70% likely happen roughly seventy times in a hundred. Being right is not the test — being right at the rate you claimed is. We measure it with the Brier score, the squared distance between the probability stated and the outcome that arrived (1 for happened, 0 for did not), the same standard used in professional forecasting tournaments: call something at 0.70 and it happens, you score 0.09; call it at 0.70 and it does not, you score 0.49; lower is better, and a coin-flipping forecaster averages 0.25. Confidently wrong is punished harder than hedged and wrong, which is precisely the incentive prophecy has never had to face. We apply it to ourselves from the first resolutions in January 2027, and from Phase 1 to every observer whose signals enter the registry.
What is falsifiability and why is it on every card?
Every projection states, in advance, the observable outcome that would prove it wrong — for example: "WRONG IF world goods-trade volume growth returns above 4%/yr for two consecutive years before 2029." If a claim cannot be tested, it does not belong on this platform.
Is this financial advice?
No, and the distinction is structural rather than a disclaimer bolted on at the end. Nothing on this platform is financial, investment, medical, legal or life advice, and nobody here is licensed to give any of it. Every forward-looking number is a probabilistic simulation, labeled as one on the card, in the chart, in the ledger row and in the footer. A 38% is not a recommendation to act; it is a statement that we expect events like this one to occur roughly thirty-eight times in a hundred, and an invitation to check that claim against us in 2027. Observed layers — the historical series behind every chart — are real published figures with their source named on the panel. If you are making an actual financial decision, take this the way you would take a weather model: as one input, not an instruction.
Where does the data come from?
Twenty tracked feeds — twelve absorbed statistical series and eight live wires. The series: FRED, World Bank Open Data, IMF WEO/COFER, BIS statistics, SIPRI military expenditure, the Caldara–Iacoviello Geopolitical Risk index, Our World in Data, the Maddison Project, Pew Research, ITU, NOAA/NASA GISS, CoinGecko. The wires: BBC, UN, NASA, ECB, Al Jazeera, NPR, the USGS earthquake feed and Wikimedia pageviews. Each source and its refresh cadence is listed in the public Model Card. Observed layers are real published figures; projections are simulations.
What are the Black Swans?
Eight low-probability, extreme-impact scenarios — engineered pandemics, cascading AI failure, solar superstorms and others — mapped with their cascade chains and early indicators. We publish them not to predict them, but so the cascade is mapped before it begins. They are balanced by Positive Futures: the measured curves bending the other way.
How do I become an Observer?
The beta is free, and registration takes ten seconds — nickname, email, age, password; no OAuth. An address gets one confirmation link — that is what attaches a purchase to your account and where resolution notices go. Signals you submit post to the PUBLIC registry, visible to every visitor; sign in to have them attributed to your handle (a local copy is kept as offline fallback). In Phase 1 (2027) public Brier-score reputation arrives, and contributed signals gain weight according to their author’s calibration history. Your calibration is your voice.
What is the long-term plan?
Phase 1 (2027): the Observer Network with reputation-weighted signals and the first public scoring. Phase 2 (2027–2028): observers pool AI API credits from any provider so the engine runs continuous multi-model scenario sweeps — where models diverge, the divergence itself becomes a signal. What comes after is deliberately unpublished until Phase 1 has produced scores worth trusting: the instrument earns its record first.
What are THE LATTICE and THE LOOM?
The August 2026 core upgrade. THE LATTICE is the platform’s knowledge as one living structure — over 200 nodes (predictions, seals, eras, swans, texts) joined by hundreds of typed links, with economic attention on top: HEAT (fast, scarce, injected by live headlines and spread by resonance flow) over WEIGHT (slow, structural). Three WARDENS keep the accounting honest, and every score opens into a public WHY ledger. THE LOOM is the persistent state layer: six threads of six named strands each, advancing every six hours on wall-clock time and live wire events — whether or not anyone is watching. Both live on /core, and the landing hero shows their live state.
How does the engine filter news into the numbers?
Headlines are ingested as raw signal only (the LATEST NEWS section) and pushed through a published 46-rule lexicon that maps wire language to bounded strand impulses on the Loom — every applied impulse is logged with its exact headline, steps are capped, and interventions are damped and publicly recorded in the LOOM LOG. Nothing a headline does can ever touch a sealed probability; it can only move the live reading around it.
What is the LIVE LEAN next to a sealed probability?
The sealed number never moves — that is the contract. The LIVE LEAN is the living instrument’s current reading on the same event: a bounded offset (at most ±7 points) recomputed from the Loom’s strands against their documented seed values, with every contributing strand listed on the prediction’s page. When the strands sit at their seeds, the lean equals the seal exactly. It is labeled a probabilistic simulation, because that is what it is.
How do the forces affect each other?
Through THE CROSSWEAVE — a published matrix of two dozen directed couplings between strands, each carrying a weight, a one-line mechanism, and the historical precedents it was calibrated against (war finance loading sovereigns: 1914, 1940, 2022; financial pain converting to institutional distrust: 1873, 1931, 2008; and so on). Disagreement with the engine becomes specific and testable, because the causal assumptions are on the record at /core.
Why should I trust these probabilities?
You should not — yet. Trust in forecasting is earned by scored resolutions, and ours have not arrived: sixteen predictions are sealed and every one is still PENDING. What you can check today is the machinery, and that is deliberately all in the open. The seal is a SHA-256 hash computed live in your browser over the ledger entries, so an edited prediction would break its own hash in public. Every projection names in advance the adjudicator that will grade it — ACLED, NASA GISTEMP and ERA5, METR, SIPRI, the G7 central banks — so we cannot pick a friendly scorer afterwards. Every model weight, source and assumption is published in the Model Card. The methodology is written out in full in How We Score Prophecy. The correct posture right now is exactly the one we ask for: withhold trust, bookmark the ledger, and come back in January 2027 when the first grades land in public whether they flatter us or not.
What happens when one of your predictions turns out wrong?
It is published as wrong, on the ledger, next to the probability we sealed, with its Brier score attached and no edit to the original number. That is the whole design — a forecasting platform that cannot record a loss is a horoscope with better typography. Two seals are already visibly at risk and we have said so on the site since August 2026: the Ukraine ceasefire seal (CAL-11, sealed at 48%) with talks stalled, and the warmest-year seal (CAL-12, sealed at 66%) with 2026 running third-warmest year-to-date. CAL-12 resolves in February 2027 and may well score INCORRECT, in public, by design. Field notes are dated annotations that track a seal as evidence moves; they never touch the sealed probability. Nothing here gets quietly deleted, and the hash over the entries makes quiet deletion detectable by anyone.
Who writes this, and what qualifies them?
One founder and an open instrument. The site is built and written by one person, working in the open, and publishes under the house name rather than a personal byline — we would rather say that plainly than invent a team or credentials. The qualification we offer is not a résumé but a standing bet: every claim is sourced, every projection is falsifiable in advance, every probability is timestamped and hash-sealed before the outcome is known, and the scoring is done by named external adjudicators on dates fixed ahead of time. That is a harder thing to fake than a biography. From July 2027 the platform carries a public Brier score, and that number — not an author bio — is the credential we intend to be judged on. Observers who contribute signals are scored on exactly the same basis: on this platform your calibration is your voice.
Is any of this written or generated by AI?
The engine is software and never pretended otherwise: the Loom advances on a published rule set every six hours, the Lattice spreads attention by a published formula, and the simulations run on a seeded generator precisely so that anyone re-running them gets the same fan, because reproducibility is part of integrity. What is not automated is the truth-checking. Every observed figure is a real published number matched to its source before it ships, historical claims are verified against primary or scholarly references, and probabilities are cross-checked against the sealed ledger so an article can never quote a number the ledger disagrees with. No content here is auto-published: nothing reaches the site without a human check against the sources. Where we do not know something, the site says so rather than generating a confident answer.
What did Nostradamus predict for 2027 and 2028?
Honestly: nothing datable. The quatrains carry almost no usable dates — the famous exceptions are the 1999 "seventh month" verse, which resolved to nothing, and the year 3797 named in the Epistle to Henry II — and every "prediction for 2027" circulating online is a modern interpreter mapping loose imagery onto a year the text never gives. What we publish instead is the calibrated substitute: twenty-eight live projections and sixteen sealed predictions covering 2027 and 2028, each with an explicit probability, a resolution date and a falsifiability condition. The long readings are in Nostradamus Predictions 2027–2030 and the 2028 calibrated reading; what the man actually wrote, and did not write, is set out in What Nostradamus Actually Predicted.
How is this different from Metaculus, Polymarket or Good Judgment Open?
Those are markets and tournaments — many forecasters aggregated into a live number that moves continuously, which is a genuinely good instrument and one we read. This is a sealed instrument, which answers a different question. Here a probability is fixed at a timestamp, hashed, and never moved again; if the live reading of the world changes we show it as a separate, clearly bounded LIVE LEAN of at most seven points beside the frozen seal, so you can always see both the call we made and the call we would make now. We publish the causal model rather than only the number: the Crossweave matrix of couplings between forces, with the historical precedents each was calibrated against, so disagreement with us becomes specific and testable instead of vibes. And the horizon is deliberately long and macro-historical — the present is scored against nine indexed stress eras from 1815 onward — which is exactly the range where liquid markets are thinnest. Nothing is traded and no money changes hands on an outcome.
How often is the site updated?
Continuously at the machine layer and deliberately never at the seal. The Loom advances every six hours on wall-clock time whether or not anyone is watching; the news wire refreshes on a fifteen-minute cache across the eight live wires; the Attention Seismograph reads Wikipedia pageviews daily; the USGS earthquake reading and the Lattice attention scores recompute on every visit to the rooms that show them. Field notes are appended to sealed entries whenever real evidence moves — the last full world sweep was 28 August 2026. Articles carry a visible UPDATED stamp and a matching dateModified when their figures are refreshed. The sealed probabilities themselves are the one thing that never updates, by contract.
Can I get told when a prediction actually resolves?
Yes, and it is the only email we send. Every prediction page, every expanded ledger row and every long article carries a resolution watch: leave an address against a specific prediction and you get exactly one message when that prediction is graded — what we sealed, what happened, and the score. No newsletter, no digest, no sharing, and deletion on request. The first resolutions land January–February 2027, and the first full public Brier review in July 2027. If you would rather hold no account at all, the ledger is public and the hash is recomputed in your browser every time you load it.
Is it free, and what do you do with my data?
The beta is free and the predictions are not paywalled — putting the calibrated numbers behind a wall would kill the scrutiny and citation that make the ledger worth anything. Registration is optional and takes ten seconds — a nickname, an email and a password, nothing else required; the address is confirmed by a link and used for the resolution alerts on the calls you seal, so the record can tell you when one of them is judged. No age gate, no third-party sign-in; user records are encrypted at rest. There is no advertising, no third-party tracking pixel and no cross-site profile — the honest banner in the corner says so because it is true, not because a regulation required a popup. Signals you submit are public by design, since a registry only means something if everyone can see it. The paid passes — a single-domain desk and the full instrument with alerts — sell tooling and standing, never access to the forecasts, which stay free.
Will AI data centers raise my electricity bill?
Probably, and it is one of the few forecasts here with a mechanism you can watch monthly. US data centers consumed 4.4% of national electricity in 2023 and Lawrence Berkeley National Laboratory projects up to 12% by 2028; the 2024 PJM capacity auction cleared at roughly ten times its prior price before the bulk of that demand had even arrived. The speed mismatch is the whole story — hyperscaler capital converts to firm 24/7 load in quarters, while transformers, turbines and transmission move at three-to-seven-year permitting speed, so the gap clears through scarcity pricing that every ratepayer helps pay. TEC-03 seals this at 66% (±10) for 2026–2029, and publishes its own kill-condition: the projection is wrong if average US retail electricity prices rise no faster than CPI through 2028, no major grid operator declares data-center-driven resource inadequacy, and interconnection queues shorten. Attention confirms the mechanism is now public property — the Wikipedia article for Data center is running about 2,345 views a day, up roughly 21% on its own prior month.
Is a stablecoin going to collapse?
Our sealed answer is that it is the least likely of the crypto-adjacent failures we price, but it is not negligible. CRY-02 seals a top-three stablecoin breaking the buck or being forced into restructuring at 30% for 2026–2029 — low, because the surviving majors are now substantially backed by short-dated Treasuries and because regulation has hardened the reserve disclosure regime; not zero, because the failure mode is a redemption run against an asset whose selling pressure hits the same T-bill market everyone else is selling into. It is graded as a real event with a date, not as a vibe, and it resolves in public alongside every other seal.
Is the dollar going to lose reserve currency status?
Eroding, not collapsing — and the distinction is where most commentary goes wrong. The published record shows the dollar’s share of allocated global reserves drifting from 71% in 1999 to roughly 58% in 2025 (IMF COFER), while central banks bought over 1,000 tonnes of gold in each of three consecutive years and the ECB reported gold overtaking the euro as the world’s second reserve asset. FIN-03 seals 45% (±14) on the share breaking below 50% by 2032. Note what the same ledger seals against itself: only 8% on the dollar falling below 45% by 2030. Reserve transitions are glacial until they are sudden — sterling’s took thirty years and one Suez — so the honest reading is erosion with a fat tail, not replacement on a schedule.
Will China invade Taiwan?
Not on the majority of our own probability mass, and we would rather say that plainly than sell the alarm. GEO-01 seals 38% (±15) on a direct US–China military incident producing casualties in the western Pacific by 2031 — a collision, shoot-down or blockade probe, not a full invasion — and the Calibration Ledger separately seals 82% (CAL-02) that no live-fire exchange of any kind occurs through 2029. The incident risk is priced seriously because Allison’s survey of power transitions found twelve of sixteen ending in war since 1500; the invasion risk is priced lower because public wargaming, most visibly the CSIS series of 2023, repeatedly finds no clean victory available to either side at costs both can see in advance, which is historically the configuration that produces pressure rather than invasion. What to watch is tempo, not headlines: monthly PLA incursion counts into Taiwan’s ADIZ, published daily by Taiwan’s defense ministry.
What does the Doomsday Clock have to do with your numbers?
It is an input we read and a format we deliberately do not copy. The Clock is a single expert-panel gauge published once a year with no probability attached and no resolution date, so it can never be scored — which is exactly the gap this platform exists to close. We track it as one slow composite signal alongside SIPRI military expenditure and the ADIZ series, and we track public attention to it too: Doomsday Clock is currently among the fastest-rising topics on our Attention Seismograph at roughly 1,279 Wikipedia views a day, up about 55% on its prior month. Where the Clock says “late”, we say 8% (±5) on battlefield use of a nuclear weapon by 2035, with the falsifiability line published beside it.
Will there be a recession in 2027?
More likely than not by the end of 2028, and the reason is the starting point rather than the trigger. FIN-02 seals 55% (±15) on a US recession beginning by end-2028 that deficits cannot cushion. The published inputs: the longest yield-curve inversion on record un-inverted in late 2024; a single benchmark revision erased 911,000 jobs from the record through March 2025; card balances above $1.2 trillion carry post-2011-high delinquency transitions; and entering a downturn with the deficit already near 6% of GDP leaves no fiscal shock absorber. One thing this number is not: a market call. Recessions and crashes are different objects, and the ledger prices them separately — CAL-04 seals 78% against the crash headline over the same window.
Are you tracking what people are searching for?
Yes, in public, and it is deliberately kept out of the projections. The Attention Seismograph on the Psyche page reads Wikipedia pageviews for ten tracked topics as a measurable proxy for collective attention, showing each topic’s trailing 30 days, its seven-day average and its ratio against the prior period. The basket is rotated against a real scan rather than intuition — the 2026-08-30 pass swept 103 candidate topics and added Data center and Gold on measured evidence. Attention is treated strictly as a signal about the crowd, never about the event: a spike is people noticing, and nothing the seismograph reads can move a sealed probability. The same discipline governs the news wire, whose 46-rule lexicon can only nudge the live reading around a seal, never the seal itself.