How we read the world.
1929.world operates as a transparent, public-source intelligence terminal. The methodology is open, the sources are cited, and the calculations are auditable. This page is the standing reference.
The premise
Two things exist in macro-financial intelligence today. Terminals (Bloomberg, Refinitiv, AlphaSense — $20–25k/seat/year) ship the data. Newsletters (Lyn Alden, Doomberg, Bridgewater Daily) ship the opinions. The seam between them — opinionated, data-backed, mobile-first, free-core, embeddable — is where we sit.
We are not a feed. We are an opinionated, transparent, investigative methodology applied to public-source data. The product is the synthesis. The proof is that the calculations are open and anyone can re-derive them.
The framework — Cui Bono, applied
Cui Bono — “to whom does the benefit accrue” — is the classical investigative question. We apply it algorithmically across multi-source public data.
Every significant event (kinetic strike, sanctions move, sovereign default, ratings change, currency intervention, leadership transition) is treated as an investigation with the structure a homicide detective would recognize: Means, Motive, Opportunity (MMO). The working assumption is universal strategic deception — no single source is trusted. Truth lives in the correlation between independent feeds.
The 5-step investigation
Every dossier renders the same five sections, in this order:
- 01Event
Ground truth from the source feed — what we observe directly. Not an inference.
Sources · ACLED · GDELT · NASA FIRMS · USGS · OpenSky · AISStream - 02Narrative
Who is publicly attributing what. Sentiment per source; semantic divergence between feeds.
Sources · GDELT (CAMEO codes) · curated headline aggregation - 03Motive
Who profits — financially or strategically? Commodity-price deltas, sovereign positioning, ETF flows.
Sources · FRED · EIA · World Bank · IMF · sovereign-profile derivatives - 04Means + Opportunity
Who had the capability and the access? UBO graphs, vessel tracks, aircraft, satellite passes, sanctions exposure.
Sources · OpenSanctions · OpenSky · AISStream · NORAD TLE · USGS - 05Propaganda Tilt
A 0–100 score: how far the loudest narrative diverges from where Cui Bono evidence points.
Sources · Derived from steps 2–4
Propaganda Tilt — the signal
Propaganda Tilt is a 0–100 score representing the distance between (a) the loudest public narrative and (b) the mathematical beneficiary surfaced by the motive + means + opportunity analysis.
| Score | Label | Meaning |
|---|---|---|
| 0 – 30 | ALIGNED | Available evidence points consistently. Multiple sources corroborate. |
| 31 – 60 | DIVERGENT | Partial alignment. Insufficient corroboration or moderate spin. |
| 61 – 85 | MANIPULATED | Loudest narrative does not match underlying evidence. Coordinated effort likely. |
| 86 – 100 | INVERTED | Loudest accusers are the mathematical beneficiaries. |
The formula is published below in full, deliberately. We do not claim infallibility — we claim auditable process. If you disagree with our reading on a specific event, you can recompute the score from the raw inputs and tell us where we went wrong.
The v1 formula
tilt(entity, sanctions) =
base 50 // agnostic default — evidence required to move
// sanctions corroboration / monosource
IF sanctions has ≥3 distinct jurisdictions AND topic="sanction"
THEN tilt -= 10 (multi-jurisdiction = corroborated)
ELSE IF sanctions has =1 jurisdiction AND topic="sanction"
THEN tilt += 15 (single-source = potentially political)
IF topics has "role.pep" AND NOT "sanction"
THEN tilt += 5 (PEP-only — context, not adverse evidence)
// geopolitical hot zone
IF entity in {ukraine, scs, taiwan, korea, hormuz, redsea}
THEN tilt += 20 (high historical narrative-manipulation rate)
ELSE IF entity in any monitored chokepoint
THEN tilt += 10 (modest baseline)
// signal/data gap
IF signal in {alert, critical} AND no enrichment data
THEN tilt += 10 (under-reported risk)
// independent indicator diversity
IF flag count ≥ 2
THEN tilt -= min(flag_count × 3, 10) (multi-source corroboration)
clamp tilt to [0, 100]
return { score, label, confidence, factors }
Source: lib/intel/cui-bono.ts. The function is a pure function with no I/O — deterministic for the given inputs. Every contributing factor renders in the dossier UI under “audit trail”.
Confidence levels
Confidence is reported alongside every score. v1 confidence is mostly low because the GDELT narrative-divergence calculation is not yet wired — we have only the structure of the score, not the full signal that makes it sharp.
- Low — current v1 default. Sanctions + flag data, no narrative analysis.
- Medium — emerges when GDELT narrative-divergence and commodity-price-windowing are wired (planned in v2).
- High — reserved for events with multiple corroborating feeds and strong narrative-vs-evidence divergence. Ships when v3 includes source-diversity weighting.
Sources we use
Everything below is publicly accessible. Most are free; ACLED has a commercial license for paid use. Where we can, we cite the upstream on every data point.
| Source | Purpose |
|---|---|
| ACLED | Strike-level conflict events, fatalities |
| GDELT | Global event database from world news, CAMEO-coded |
| NASA FIRMS | Satellite-detected thermal anomalies |
| USGS Earthquakes | Real-time seismic events |
| OpenSky Network | ADS-B aircraft positions |
| AISStream | Live AIS vessel tracks |
| OpenSanctions | PEP / sanctions / watchlist enrichment |
| FRED (St. Louis Fed) | Macro indicators, rates |
| EIA | Crude oil, gas storage, electricity |
| NOAA SWPC | Space weather, geomagnetic alerts |
| NORAD TLE (CelesTrak) | Two-line-element satellite catalog |
| World Bank Open Data | Sovereign indicators, debt, demographics |
CAI versus classified — what we do not have
We operate entirely in Commercially Available Information (CAI) — the same category of public data the ODNI’s 2022 report acknowledged is reshaping the intelligence landscape. We do not have access to classified SIGINT, HUMINT, or imagery. What we can do is cross-correlate public sources with the same rigor an analyst would inside a classified system.
Where intelligence agencies pay $50,000–$500,000 per seat per year for Sayari, Babel Street, Blackbird AI, and similar enterprise platforms, we ship the equivalent analytical structure on free public APIs. Our edge is not access — it is design and synthesis.
Known limitations of v1
- Single-event scope. The score evaluates one event at a time. A multi-event correlation engine (pipeline sabotage → LNG spread → SOE shipping pattern) is v2.
- No live narrative analysis. GDELT integration is not yet wired. The Narrative step shows attribution from source feeds only; the divergence calculation is a placeholder until the GDELT pipeline lands.
- Coarse motive heuristic. Regional motive context is currently rule-based (Hormuz → Brent, Suez → tanker rates). Live commodity-price-windowing relative to the event timestamp is v2.
- English-language bias. Our headline aggregation skews to English-language sources. Babel-Street-style multilingual deep-web monitoring is out of scope.
How to audit us
- Open any entity in the /ops workbench.
- Toggle the dossier panel to CUI BONO mode.
- Expand the audit trail on the Propaganda Tilt section. Every contributing factor is listed with its delta and reasoning.
- Cross-check against the source feeds listed above. If you find a factor that should not have applied, or a factor we missed, open an issue with the entity ID.