Chokepoint Data

Fast sources refresh hourly. Brent crude (FRED), prediction-market odds, Iranian state-media statements (IRNA, Tasnim, Mehr), and the public Truth Social archive are pulled every hour and every number on the site recomputes from them. The headline Brent price (marked ≈) and its % since closure are an indicative futures quote refreshed on the same hourly cycle and shown with its trade timestamp; official EIA daily closes — which publish several days late — remain the basis for every other computed statistic: the analysis, the kill gate, and the charts. Transit counts refresh weekly — IMF PortWatch publishes Tuesday mornings (snapshotted, so their retroactive revisions stay visible), for Hormuz and for the four chokepoints the timeline compares it against: Suez, Bab el-Mandeb, Malacca and the Cape of Good Hope, each indexed to its own pre-Houthi average so the two reroutings can be read side by side. The cost-of-transit measures on the timeline — VLCC freight out of the Gulf, the Brent-over-WTI premium, the futures curve, oil-implied volatility, the Tehran rial — are collected hourly too; each appears as a panel there once it has enough history to draw honestly. Each refresh commits and the site deploys itself. Nothing is edited by hand between the raw data and what you read.

Breaking events, by hand. Major events are curated and source-verified manually and pushed as they happen, which reruns the same computation — that push is what feeds the Atom feed and subscriber alerts.

Daily analysis, with a kill gate. Once a day an automated analyst checks whether the data actually moved — a new major event, an odds swing, an outsized Brent move, a burst of statements. If nothing cleared the bar, it writes nothing; most days should be silent. When it does write, every number is computed from the same JSON the charts read, and the piece publishes directly — review happens in the open, after the fact, and pieces that miss the bar get removed. When a piece publishes, subscribers whose stated preference matches it get one email linking here.

Freshness is visible. The footer on every page states when the site was last built and the exact date each dataset runs through. If a number looks stale, that's why — we'd rather show a dated truth than a fresh guess.

Everything on this site is computed from JSON you can fetch directly — no key, no rate-limit games. If you're wiring up an agent (a Claude Code or Codex job, a cron, a GitHub Action) to react when something breaking happens here:

Poll the feed. feed.xml is an Atom feed with one entry per major event and analysis piece, stable ids for deduping, and computed reactions in the summary. New entry → trigger your job.

Read the data. The full datasets behind the charts: timeline.json (Brent, transits, events with computed reactions), markets.json (prediction-market series), incidents.json, iran_statements.json, truths.json, analysis.json and calibration.json. Updated with each data refresh; please keep the source citations if you republish.