Brújula Systems
Home What is The Model Editorial Personal ES
The model · updated every 30 minutes

Where the world is moving

8 master theses. 39 granular sub-theses. Anchored to public, verifiable sources. Probabilities move as evidence arrives.

8 master · 39 sub-theses · updated today ▲ up ▼ down
The eight big questions

Each sub-thesis says what has to happen for it to count as resolved. Only facts after 7 May 2026 count — the day the system started watching: a thesis whose criterion was already met before then is reformulated, not resolved.

TM-1 · Acceleration and Agentic Frontier 61%

7 sub-theses

  • IA-1.1 · AGI before 2030 45%
    What has to happen: A system with cognitive capability equivalent to or greater than human in most economically relevant tasks, documented by at least two independent public evaluations (e.g. METR, Epoch AI, AISI), before 2030-12-31.
  • IA-1.2 · Frontier AI concentration in ≤3 actors 40%
    What has to happen: ≥80% of the productive compute of frontier models (>1e26 FLOPS) in 3 or fewer organizations.
  • IA-1.3 · Autonomous agentic capability: p50 time-horizon ≥1 month <2027 55%
    What has to happen: METR publishes an eval in which a frontier model reaches p50_horizon_minutes ≥ 43,800 (~1 month) on HCAST.
  • IA-1.4 · AI capability/investment plateau before 2028 6%
    What has to happen: Frontier AI improvement flattens: Elo ceiling on LMArena stagnant ≥6 months, or a year-on-year drop of at least 25% in estimated frontier model training spend according to Epoch AI (or in the training compute of the largest model of the year), across three consecutive readings of the series, or deceleration documented by METR/AISI.
  • IA-1.5 · High-profile alignment failure <2028 60%
    What has to happen: One of: (a) deployed model causes measurable physical/financial harm documented by an independent source; (b) public revocation or recall due to alignment risk; (c) AISI or equivalent publishes a formal warning.
  • IA-1.6 · Documented loss of control of AI systems before 2028 84%
    What has to happen: Public and verifiable incident in which a system with agency (agents, robots, trading) falls outside operational control and acts against its specification with material consequences. Not a chatbot that goes off script.
  • IA-1.8 · Compute/datacenter bottleneck slows AI frontier expansion before 2028 80%
    What has to happen: Verifiable public statement from a frontier lab announcing the delay or cancellation of a large project due to energy/chip/interconnection access, or an interconnection queue sustained >5 years in PJM, ERCOT or MISO (editorial verification; does not move the probability on its own).

TM-2 · Geoeconomic and Financial Fracture 30%

5 sub-theses

  • GEO-2.1 · Military conflict in Taiwan Strait before 2030 45%
    What has to happen: Open military action by the PRC against Taiwan —blockade, invasion or sustained kinetic exchange—.
  • GEO-2.2 · Financial multipolarity: USD share in global reserves below 50% before 2030 30%
    What has to happen: The dollar's share of official reserves according to the IMF COFER falls below 50%.
  • GEO-2.3 · Major OECD fiscal crisis <2030 40%
    What has to happen: A G7 economy with (a) debt/GDP >150% and (b) sovereign spread >400 bp over the US 10Y for 60 consecutive days; or (c) an IMF or ECB bailout.
  • GEO-2.4 · Sovereign Bitcoin reserve adoption 9%
    What has to happen: ≥5 G20 countries with an allocation >0.5% of the sovereign balance sheet in BTC.
  • GEO-2.7 · Operational use of nuclear weapon before 2030 23%
    What has to happen: Visible atmospheric test, demonstrative detonation or actual attack by any nuclear State. Excludes routine underground maintenance tests.

TM-3 · Labor Disruption and Value Capture 49%

4 sub-theses

  • ECO-3.1 · Mass white-collar labor disruption <2028 57%
    What has to happen: One of: (a) sustained >15% drop in junior administrative/legal/financial employment in the G7 according to BLS/Eurostat; (b) ≥3 top-100 firms announce layoffs >50% of staff explicitly attributed to AI; (c) white-collar output per worker breaks the historical trend by >2 standard deviations.
  • ECO-3.2 · Automation rent capture: employment and wages both fall in information and professional services before 2030 53%
    What has to happen: (a) ≥12-month drop in employment in NAICS 51+54+55 >2% year-over-year with wages also falling; or (b) BLS output per hour rises >5% year-over-year while headcount in the same sector falls.
  • ECO-3.3 · 100k+ humanoid robots operational in production before 2028 39%
    What has to happen: The global fleet of humanoids deployed in industrial production, logistics or services exceeds 100,000 operational units, not prototypes or demos.
  • ECO-3.5 · AI risk securitization / liquid p(doom) markets 41%
    What has to happen: One of: (a) a prediction market on catastrophic AI risk with volume >1 M USD sustained 90 days; (b) formal issuance of derivatives on material AI harm on a regulated exchange; (c) FT or WSJ treat «p(doom) markets» as an established category.

TM-4 · Physical-Infrastructural Bottleneck 61%

6 sub-theses

  • TEC-4.1 · Sustained net-positive commercial fusion <2035 46%
    What has to happen: A reactor reaches sustained Q>1 in commercial operation (not laboratory), or a consortium announces a first commercial plant with a binding date.
  • TEC-4.2 · Five or more G20 countries with >50% renewable electricity before 2030 64%
    What has to happen: ≥5 G20 members sustain renewable generation (solar + wind + hydro + geothermal + bioenergy) above 50% for two consecutive years. Nuclear does not count.
  • TEC-4.3 · Electric bottleneck slows AI datacenter expansion before 2028 93%
    Resolved on 2026-07-14 · happened · last forecast 93%
    What has to happen: Announced cancellation due to lack of grid access, regional moratorium, or connection auction with a wait >5 years in PJM, ERCOT, MISO, SPP or CAISO (this last one, editorial verification; does not move the probability on its own). Also a z-score >2σ in daily demand sustained for three consecutive days in the same region.
  • TEC-4.4 · Commercial SMR reactor connected to the grid before 2030 52%
    What has to happen: An SMR in commercial operation —generation and continuous dispatch to the public grid, not a demo— of a replicable Western design, or a new Asian model with a commercial order outside its country.
  • TEC-4.5 · Quantum Q-Day before 2032 74%
    What has to happen: A quantum system demonstrates at least 100 logical qubits with error correction operating below threshold, in a peer-reviewed publication or independent verification, before 2032-12-31.
  • TEC-4.6 · Critical semiconductor disruption <2028 70%
    What has to happen: One of, with a date after 2026-01-01: (a) sanction or blockade of TSMC >30 days due to a geopolitical event; (b) interruption of ASML's EUV chain >60 days; (c) a new sanction that prohibits a major actor from accessing a node ≤3 nm. Restrictions in force before 2026 do not count.

TM-5 · Biological and Transhumanist Frontier 46%

5 sub-theses

  • BIO-5.1 · Anti-aging therapy regulatory approval <2030 13%
    What has to happen: FDA or EMA approves a therapy whose official indication explicitly mentions «aging» or «senescence» as the primary target, not just associated disease.
  • BIO-5.2 · Systemic-impact biosecurity incident before 2030 29%
    What has to happen: Laboratory leak with technical attribution, deliberate use of a biological weapon, or outbreak of an AI-designed pathogen, with >10,000 confirmed victims or sectoral collapse plus major regulatory response.
  • BIO-5.3 · Modular CRISPR platform approved <2030 52%
    What has to happen: FDA or EMA approves a CRISPR therapy whose regulatory documentation describes the platform as «modular» or «programmable» for multiple indications, not a single monogenic one.
  • BIO-5.4 · Commercial brain-computer interface in production before 2030 27%
    What has to happen: A BCI device with FDA approval and commercial deployment —not a trial, not compassionate use— with more than 1,000 patients implanted.
  • BIO-5.6 · AI-designed protein leading to approved therapy <2030 90%
    What has to happen: A therapy designed de novo by AI systems reaches Phase II in humans or FDA approval.

TM-6 · Demographic and Anthropological Reconfiguration 55%

4 sub-theses

  • SOC-6.1 · Asymmetric demographic collapse OECD+China 60%
    What has to happen: ≥3 OECD countries + China with (a) TFR sustained <1.3 for ≥10 consecutive years, or (b) annual population decline >0.5% without migratory compensation, or (c) dependency ratio (65+/15-64) >50% according to UN DESA.
  • SOC-6.2 · 20M+ people displaced by climate or conflict before 2030 94%
    Resolved on 2026-05-12 · happened · last forecast 94%
    What has to happen: Net flow of people forcibly displaced by acute climate causes or conflict reaches ≥20 million in any 12-month window, per consolidated UNHCR + IDMC figures (editorial verification; does not move the probability on its own).
  • SOC-6.3 · Terminal polarization of OECD democracies <2032 34%
    What has to happen: A G7+ democracy with two of three: (a) >40% of the electorate prefers a non-democratic system, sustained 24 months; (b) institutional rejection of an electoral result by the executive; (c) political violence with >100 deaths linked to electoral tension.
  • SOC-6.4 · Terminal institutional crisis — sustained trust <30% 50%
    What has to happen: ≥3 of the 5 largest OECD economies (US, UK, DE, FR, JP) with <30% public trust in ≥3 core institutions, sustained 24 months, according to Edelman or Eurobarometer.

TM-7 · Civilizational Crisis of Meaning and Purpose 68%

3 sub-theses

  • SEN-7.1 · Collapse of intergenerational optimism 85%
    What has to happen: ≥5 G20 democracies where <30% of those under 35 expect the next generation to live better, sustained 24 months according to Pew, Edelman or Eurobarometer.
  • SEN-7.5 · Measurable collapse of shared public epistemology 50%
    What has to happen: Two of: (a) the dispersion of factual beliefs in the OECD exceeds its historical threshold (editorial verification; it does not move the probability on its own); (b) Edelman Trust in information <30% in ≥5 major democracies, 24 months; (c) >40% of information consumption in a major democracy comes from synthetic media without verification (editorial verification; it does not move the probability on its own); (d) documented collapse of a major verification institution.
  • SEN-7.6 · Unverifiable synthetic content overtakes verifiable content before 2028 70%
    What has to happen: It is publicly documented that >50% of new content on the major platforms is generated by AI without a verifiable provenance mark (C2PA, Content Credentials, watermarking), according to NewsGuard, Cloudflare Radar or a peer-reviewed study.

TM-8 · Stability of Planetary Biophysical Systems 54%

5 sub-theses

  • BIO-X.1 · Identified climate tipping point crossed before 2035 41%
    What has to happen: Verifiable crossing of ≥1 IPCC AR6/AR7 tipping point: Amazon dieback (>40% of biomass), AMOC weakening (>30% vs. preindustrial), Greenland (>500 Gt/year sustained), Arctic permafrost (>5% of area), or reefs (>90% global mortality).
  • BIO-X.2 · Regional water crisis with major geopolitical impact before 2030 70%
    What has to happen: Before 2030-12-31, in one of the five basins (Indo, Nile, Mekong, Colorado/US Southwest, Andes): UCDP records an armed conflict (≥25 combat deaths in a year) between riparian States with water as the declared cause, or a riparian State suspends or formally denounces an existing water treaty.
  • BIO-X.3 · Fisheries or agricultural collapse in key producing region before 2030 41%
    What has to happen: A world top-10 region for cereals, oilseeds or fisheries with (a) a >25% drop in annual production according to FAO/USDA, or (b) FAO Food Price Index sustained >150 for ≥12 months with a cascade effect in an OECD net importer.
  • BIO-X.4 · Biodiversity cascade with measurable economic impact before 2032 54%
    What has to happen: (a) Global Living Planet Index below 25 in a WWF/ZSL publication —the regional LPI does not count—, or (b) IPBES publishes an assessment with a loss >5 trillion USD/year in ecosystem services, or (c) FAO documents productive collapse in >2 G20 countries at once due to pollinator loss. Clause (b) is verified editorially; it does not move the probability on its own.
  • BIO-X.5 · Record climate displacement: 50 million disaster displacements in one year before 2032 66%
    What has to happen: IDMC records 50 million or more internal displacements due to disasters in a single calendar year, before 2032-12-31 (editorial verification; it does not move the probability on its own).

Model state as of 2026-11-03 03:14 UTC · This copy is what you see when printing the page or reading it without JavaScript. With JavaScript, the numbers are fetched live on load.

How probabilities move

01
Each sub-thesis is anchored to public sources: prediction markets, papers, regulatory registers, macro indicators and biophysical observations.
02
Every relevant event runs through an extraction layer that assigns magnitude and direction. Movements over the editorial threshold pause for human validation.
03
Master theses aggregate their granulars via weighted average: high confidence weighs 3, medium 2, low 1.

Two further rules correct known biases, and both are automatic: a sub-thesis that goes more than 30 days without a signal drifts a little closer to its base rate each day, because nothing happening is information too; and a source that in 30 days almost never contradicts any thesis adds at most three points to a thesis; what is left over is recorded and not applied, because a source that only confirms is not evidence, it is a sampling bias.

What this is not

It is not a forecast. It is not a recommendation. It does not replace your judgment. The model is an instrument: it shows the structural state of variables that compound across decision horizons.

Published predictions carry dates and remain public. Annual Brier-score audit against forecasting baselines.

⌖ © 2026 · Building in public Privacy Terms
Read the editorial archive →