The Daily Mayo
How this works

What is measured, and what is not

Three honest limits, stated up front, then what the grades mean.

The three limits

This is a record of what one mining run saw, not a record of the world. A story absent here was either not published, not linked by a machine-readable source, or filtered out. Absence is never evidence.

The grades describe provenance, not truth. A government advisory scores high because it is the document, and a document can still be wrong, late, or self-serving. A newsroom account scores lower because it is a description of a document, even when the description is excellent.

The source statistics are self-referential: they measure this archive's own behaviour. A source can show a low share because the brief rarely cites it, which is a fact about the brief, not about the source.

The grades

Primary record. The advisory, filing, register or mailing-list post that the claim is about. Cite it directly.

Source artifact. The repository, release, paper or dataset itself. The file is the record.

Reporting. A newsroom account written by someone who read the record. Reliable in proportion to how much of the underlying document it shows.

Aggregator. A pointer to someone else's page. The grade follows the destination.

Commentary. A reading of events by a practitioner or writer. Useful, unfalsifiable without the underlying facts.

Unattributed. Surfaced with no link to any record at all. The weakest entry in the archive.

Statements are not facts

When a vendor, government or company says something happened, this archive says they said it. That is a statement, and it is labelled as one. Turning statements into facts requires independent evidence, which is the layer still being built.

Paywalls and the archive

Where a publisher hides the text, the free copy is linked instead: a web-archive snapshot, or the archived copy taken at reading time. Every item keeps a local snapshot when one could be captured, so a dead link or a moved paywall does not erase the record.

What would make this better

Claim-level verification: for each factual assertion in an article, whether a second independent source confirms it, contradicts it or stays silent. That has to accumulate from outcomes over time. Any score invented before that evidence exists would be decoration.

Last built . Data and presentation are separate: every item is a JSON record, so the design can change without touching the archive.