How to read it: This is coverage, not the market. A tile appears only where we have a sector, a market cap and two recent bars — so an absent sector means we have no fresh price there, not that it was quiet. And the blocks are our SIC crosswalk, not GICS: Alphabet and Meta are drawn inside Technology, Visa inside Industrials, UnitedHealth inside Financials. The Technology rectangle is therefore not the Technology sector. Details below.

How to read it: This is a snapshot, not a live tape. The page re-polls every 30 seconds, but the prices behind it are cached and refreshed about every two minutes — the stamp above the table tells you how old they actually are. Past 24 hours we stop serving them altogether rather than render a stale price as a live one: that is CoinGecko’s licence condition, and it is also just the right thing to do.

How to read it: Nothing on this tab is painted green for going up. A rising inflation or unemployment print is not a “good” number, so the deltas here are deliberately colourless — the arrow gives direction and the colour claims nothing. Each change states its unit (index points are not percent) and the period it spans, because they differ: the yields step one business day while CPI and unemployment step a month. “Fed Funds Target” is the actual daily policy rate; “Fed Funds (eff.)” beside it is last month’s average and lags the Fed by design. In the positioning table, the percentile is against each market’s own stored history — at most three years and often less — and the window is printed beside every number.

How to read it: the United States has no county-level or station-level retail fuel price. None is published, and none is available to buy on terms we can use. EIA surveys roughly 900 outlets nationally by telephone and reports 29 geographies: the nation, 9 PADD regions, 9 states and 10 marketing cities. Every other area on this map shows an inherited price.

Strategy Backtester · single-ticker, no-lookahead, with the full honesty stack (research lens)

Intraday · IEX minute bars — the chart day traders watch, plus a strategy backtester (research)

Strategy backtester · flat-at-close, no-lookahead — research, not a signal

How to read it: A blank is missing data, not a zero — hover any dash and it will tell you why (a blank Div Yld does not mean the company pays nothing). Hover a value to see its vintage: some cells are a vendor’s trailing-twelve-month figure and some are computed from the last annual filing, so sorting a column partly sorts by data age. Market caps are captured per-symbol at each ticker’s last enrichment, so the rows are not priced at the same moment — the spread is printed under the filters. “Quality” is 40% profitability, 35% balance sheet and 25% valuation, so an excellent business that is expensive is marked down; a score built from fewer than three legs is marked with a dot and is not on the same scale as one built from three.

Factor Lab · point-in-time factor backtest (research lens)

How to read it: The Company column shows the filing’s SUBJECT first and its co-filers after the “+”. This matters more than it sounds: on a Schedule 13D, EDGAR names the target issuer first and the activist second — so a search for a fund returns rows headed by the companies it is targeting, and the fund itself appears after the plus. Both date boxes bound the search; fill only one and the other is completed against the corpus edge, so the filter does what the box says.

How to read it: These are disclosed facts, not forecasts. A name appearing here means a filing said something; it does not mean the price will move — the event studies we have run against these signals came back a robust null, and we would rather publish that than hide it. “Groups” counts distinct 13D reporting groups, not co-filers, so one fund filing with twelve affiliates counts once. Convergence counts distinct evidence, so two screens reading the same 8-K count once. Every card states its own coverage window — and an empty card is information: it means the screen genuinely found nothing, not that it is broken.

How to read it: “Top” ranks by relevance, not recency — importance × section × a 36-hour decay — so the lead story can be a day old and still be the lead. “Latest” is the chronological feed. “High impact” is a severity class stamped at filing time, not a claim that something is breaking: only items filed in the last six hours are badged Breaking. Every press headline is classed medium, so “High impact only” narrows the feed to primary sources.

Real-world headlines discovered via The GDELT Project — each links to the original publisher; article text is not reproduced. Gov & SEC content is public domain.

How to read it: “Last Job Runs” is how long ago a job last STAMPED a run, which is not the same as how long ago it was healthy: a job that self-bails because its freshness marker is still valid returns without stamping, so a perfectly fine hourly job can show an old time. The Errors card under-reports — several jobs log warnings without writing an error key, so an empty card is not a clean bill of health. In the cross-check, a source that has stopped answering is reported as exactly that, and kept out of the divergence number: the gap between a dead tick and a live price is elapsed time, not disagreement.