> ## Documentation Index > Fetch the complete documentation index at: https://docs.gdeltcloud.com/llms.txt > Use this file to discover all available pages before exploring further. # Atlas Vane > Entity- and question-conditioned readings — exposure, risk, and stance for a specific company or actor rather than a place, anchored where possible to real prediction-market prices. **Planned.** Vane is designed but not built. This page states the intent and the influences. Nothing here is available on the API yet. ## What it is Every other Atlas layer indexes a **place**. **Atlas Vane** indexes an **entity or a question**. The name is the point: a weathervane shows which way the wind is blowing for a specific thing. Vane answers questions of the form *what is the geopolitical exposure of this company*, or *which way is this situation trending*, rather than *how tense is this country*. ## Why you should care **Country risk is a poor proxy for entity risk.** A firm's exposure is determined by where its facilities, counterparties, shipping routes, and regulatory relationships actually are — not by its country of domicile. A company headquartered in a calm country can be almost entirely exposed to a volatile one, and a country-level index will never show it. **We already hold the join.** GDELT Cloud resolves entities across news, corporate filings, energy and industrial assets, maritime movements, government awards, and screening lists. Vane is the layer that turns that join into a reading — entity exposure computed through the assets and relationships we can actually evidence, rather than inferred from a headquarters address. **Anchoring beats asserting.** Where a liquid prediction market exists for a geopolitical question, a real traded price is a better calibration anchor than a model's own confidence. Vane is intended to use on-chain **Polymarket** prices as that anchor where they exist, and to be explicit when they do not. ## What it is useful for * **Portfolio and counterparty exposure.** Which holdings are actually exposed to a developing situation, via which assets. * **Supply-chain risk.** Exposure through facilities and shipping rather than through domicile. * **Situation tracking.** A directional reading on a specific question, with the evidence attached. * **Calibration.** Comparing a model-derived reading against a market price for the same question. ## Design intent * **Evidence-bundled.** Every reading ships with the specific events, filings, assets, and sources behind it. A reading you cannot audit is not useful for a risk memo. * **Explicitly not for screening.** Vane is an analytical exposure reading, **not** a sanctions or compliance determination, and will be labeled as such. Compliance decisions belong to the screening surfaces, which are built for that purpose and carry the appropriate provenance. * **Market-anchored where possible.** Polymarket on-chain prices as the calibration anchor for questions that have one. This was a deliberate choice over other venues for on-chain verifiability. * **Null where unevidenced.** As everywhere in Atlas, an entity with insufficient linkage returns nothing rather than a confident-looking default. ## Influences and references On-chain prediction markets. Intended as the price anchor for question-conditioned readings, chosen for on-chain verifiability of settlement and price history. The forecasting-tournament literature establishing that well-calibrated probabilistic judgment is measurable, improvable, and best evaluated with proper scoring rules — the standard any Vane reading should be held to. "Firm-Level Political Risk: Measurement and Effects," *Quarterly Journal of Economics* 134(4), 2019. Demonstrates that political risk is meaningfully firm-specific and measurable from text — the academic grounding for indexing entities rather than places. ## Status Not started, and sequenced after Posture and CNP. The entity-resolution infrastructure Vane depends on is live today and used by the existing entity surfaces.