// methodology
How l0g reads risk
l0g treats risk as a set of observable pressures, not as a prophecy. The method is simple: use primary sources, keep dates explicit, expose uncertainty, separate signal families and look for confluence before drawing conclusions.
1. Observe the source
Start from the original dataset, filing, release, regulator note or official time series whenever possible. Secondary coverage is useful for context, not as the root of a claim.
2. Separate signal families
A risk score is not a truth machine. l0g separates macro, liquidity, credit, market plumbing, policy, sanctions and crypto-regulatory signals instead of collapsing them into a magic number.
3. Preserve time
Signals keep publication dates, observation dates, retrieval dates and snapshot hashes where available. Backtests must use what was observable at the time.
4. Expose uncertainty
Coverage gaps, stale fields, imputed buckets and model caveats are part of the output. Hidden uncertainty is worse than visible incompleteness.
5. Prefer confluence
A single metric is rarely enough. l0g looks for alignment between independent signals: data, filings, positioning, funding stress, official releases and market prices.
6. Refuse false precision
A score can help compare regimes, but decimals do not make uncertainty disappear. The site avoids claims that a dashboard cannot support.
How dashboards should be read
Dashboards normalize heterogeneous signals so they can be compared visually. That normalization is a reading aid, not a unified physical measure. When a score is missing source coverage or relies on neutral imputation, the coverage itself matters as much as the number.
The machine-readable layer keeps this constraint visible through fields such as freshness, coverage, policy, scale caveats and precision guards.
How evidence should be cited
A claim should carry its source, URL, date and status: observed fact, estimate, inference, scenario or uncategorized assertion. The Agent Surface and the evidence graph exist to preserve that chain when agents reuse l0g.