Benchmark

Facts vs Forecast vs Intelligence

Why not just read a forecast or an intelligence feed? Here is the same world-topic through three columns — azimuth (observed facts from the live 2026-W35 bundle, every claim clickable to its L1 source), a forecast product (a model probability), and an intelligence product (an analyst assessment). A fair contrast, not a strawman: azimuth wins on provenance, neutrality and reproducibility; a forecast / intel feed legitimately wins on forward-looking coverage. The forecast / intelligence columns quote WorldMonitor as the compared product (captured 2026-08-24T07:16:10Z), deliberately not a clickable L1 link — because that is exactly the difference.

EU / global energy supply security

Head-to-head: Is the world's oil & gas supply getting safer or more fragile?

azimuth — observed facts

EU gas storageUS crude inventoriesEnergy prices

EU gas storage stands at 3,169 Bcf (+16 Bcf w/w) as of 2026-08-14 — an observed reading (natural-gas-storage-eu)
US crude inventories drew down -863 to 722,241 Mb (EIA week of 2026-08-14) — an observed reading (crude-oil-inventories)
Spot crude as reported: WTI $84.05/bbl (+6.5 w/w), Brent $92.51/bbl (+5.3 w/w) — the observed price tape (energy-prices)

Forecast product — projection

Assigns a 40% probability over 30d that “Oil price impact from Strait of Hormuz disruption” — confidence 58%, trend stable.

Probability path: d30=17%, d7=23%, h24=40%

WorldMonitor forecast feed (model projection), captured 2026-08-24T07:16:10Z

Intelligence product — assessment

Strait of Hormuz risk: critical is setting the strategic baseline, and the current 40% probability implies a live but not settled risk path.

Actor lens: Commodity traders: Price whether stress in Middle East becomes durable over the 30d. Rebalance positions if the probability path moves away from 40%.

WorldMonitor intelligence assessment (analyst judgement), captured 2026-08-24T07:16:10Z

Dimension	azimuth	Forecast	Intelligence
Every claim sourced?	Yes — each claim links to its L1 note	No — the probability is asserted, not sourced	No — the assessment is asserted, not sourced
Observed vs predicted?	Observed — reports what already happened	Predicted — a probability over a horizon	Assessed — an analyst judgement
Neutral vs takes a position?	Neutral — states the signal, takes no side	Takes a position — it is a forward call	Takes a position — actor lenses imply action
Reproducible / regenerable?	Yes — rerun the open-data pipeline, same result	No — a model output you cannot rebuild	No — a judgement you cannot rebuild
Free + open-licensed?	Yes — CC-BY-4.0, public-domain sources	Compared product (proprietary model output)	Compared product (proprietary model output)
Forward-looking coverage?	No — azimuth is a rear/now view, not a crystal ball	Yes — this is its job: it looks ahead	Yes — it offers a forward assessment

Trust test: the forecast tells you a probability of oil-supply stress; azimuth tells you what the physical balances and the price tape ACTUALLY did this week, and lets you click each figure back to its source. The forecast is the better crystal ball; azimuth is the better witness — provenance you can verify, neutral wording, and a result you can regenerate yourself (crude-oil-inventories, energy-prices)

A maritime / supply-chain shift and its inflation pressure

Head-to-head: Is a shipping/supply-chain shock feeding inflation pressure?

azimuth — observed facts

Energy pricesUS crude inventories

WTI crude as reported: $84.05/bbl (+6.5 w/w) — the observed energy-cost input to inflation, not a forecast of it (energy-prices)
US crude stocks -863 (week of 2026-08-14) — the observed physical balance behind the cost line, every figure clickable (crude-oil-inventories)

Forecast product — projection

Assigns a 66% probability over 30d that “Sovereign risk repricing from Black Sea maritime disruption state” — confidence 75%, trend stable.

Probability path: d30=28%, d7=38%, h24=66%

WorldMonitor forecast feed (model projection), captured 2026-08-24T07:16:10Z

Intelligence product — assessment

Black Sea maritime disruption state is keeping sovereign risk elevated through sovereign stress, with 78% pressure and 72% state confirmation is setting the strategic baseline, and the current 66% probability implies a live but not settled risk path.

Actor lens: Commodity traders: Price whether stress in Black Sea becomes durable over the 30d. Rebalance positions if the probability path moves away from 66%.

WorldMonitor intelligence assessment (analyst judgement), captured 2026-08-24T07:16:10Z

Dimension	azimuth	Forecast	Intelligence
Every claim sourced?	Yes — each claim links to its L1 note	No — the probability is asserted, not sourced	No — the assessment is asserted, not sourced
Observed vs predicted?	Observed — reports what already happened	Predicted — a probability over a horizon	Assessed — an analyst judgement
Neutral vs takes a position?	Neutral — states the signal, takes no side	Takes a position — it is a forward call	Takes a position — actor lenses imply action
Reproducible / regenerable?	Yes — rerun the open-data pipeline, same result	No — a model output you cannot rebuild	No — a judgement you cannot rebuild
Free + open-licensed?	Yes — CC-BY-4.0, public-domain sources	Compared product (proprietary model output)	Compared product (proprietary model output)
Forward-looking coverage?	No — azimuth is a rear/now view, not a crystal ball	Yes — this is its job: it looks ahead	Yes — it offers a forward assessment

Trust test: the forecast assigns a probability to a maritime-disruption / inflation path; azimuth does not predict the disruption — it surfaces the observed energy-cost facts it CAN see, each linked to source. Read the forecast for the forward call; read azimuth for the verifiable, neutral baseline it is told against (energy-prices, crude-oil-inventories)

A seismic event near infrastructure

Head-to-head: Did a major earthquake this week threaten energy infrastructure?

azimuth — observed facts

Geophysical

Largest recorded event this week: M6.7 31 km NW of Aniso, Peru — one of 48 events at or above M5 USGS logged, an observed record (earthquakes)
azimuth reports what USGS RECORDED, never what may happen next — a sourced, neutral record of seismicity, regenerable from the open USGS feed (earthquakes)

Forecast product — projection

No product publishes a seismic forecast — earthquakes are not predictable, so there is no forecast/intel alternative to azimuth's observed USGS record. This is the honest inverse: for an observed natural-hazard fact, the sourced record IS the only trustworthy source.

Intelligence product — assessment

Dimension	azimuth	Forecast	Intelligence
Every claim sourced?	Yes — each claim links to its L1 note	—	—
Observed vs predicted?	Observed — reports what already happened	—	—
Neutral vs takes a position?	Neutral — states the signal, takes no side	—	—
Reproducible / regenerable?	Yes — rerun the open-data pipeline, same result	—	—
Free + open-licensed?	Yes — CC-BY-4.0, public-domain sources	—	—
Forward-looking coverage?	No — azimuth is a rear/now view, not a crystal ball	—	—

Trust test: nobody credibly forecasts earthquakes, so there is no forecast/intel column to compare — the observed USGS record IS the only trustworthy source, and it is exactly azimuth's lane: sourced, neutral, regenerable. The honest limit cuts the other way too — azimuth tells you what shook, never what will (earthquakes)

How this benchmark is made

azimuth’s columns are generated deterministically by synthesis/benchmark.py from the live L1 bundle — every figure read straight from a source note, every claim carrying its inline L1-source citation, nothing invented and nothing predicted. The forecast and intelligence columns are a dated snapshot of the compared product, captured by scripts/pull_benchmark_foils.py on the weekly cadence and quoted by product + date. azimuth’s own columns never editorialise or predict.