Methodology

This page describes how records get onto the site. It describes the pipeline as it runs, not an aspiration.

1. Source monitoring

The pipeline watches a registry of sources: RSS feeds, company pages, news indexes, video channels, and public social feeds. Each item the watcher picks up keeps a link to where it was found.

2. Scoring

Detected items are scored for newsworthiness on a 0 to 100 scale. The rubric weighs novelty, magnitude, visual potential, freshness, and fit for the audience. High scores move fast, low scores wait or drop.

3. Verification

Publication runs on two lanes. Records from sources on an autopublish allowlist, or picked in an operator-run batch, go to an independent model checker; when the check passes, they publish automatically with any uncorroborated high-stakes figure (amounts, valuations, specifications) held back. Every other pipeline-detected record queues for a human curator, whose review is the check for that record. In both lanes, whoever drafted the record never checks it: the maker is never the checker.

These gates landed in mid-July 2026. Records published before then predate them and were not individually re-verified; a correction to one of those records goes through the gates like anything else.

4. Publish gates

A record that cannot name its source does not publish. Gates also check risk and structure before anything goes live.

Claim classes

Every claim on the site carries a class that says who stands behind it:

ClassMeaning
manufacturer-claimedThe company that builds the robot said it.
supplier-claimedA supplier said it, usually in its own disclosures or filings.
third-party-measuredSomeone outside the companies involved measured it.
regulator-filedThe figure comes from a filing with a regulator.
independently-verifiedThe claim was confirmed against sources independent of the claimant.

The class is the honest answer to “how do you know”. A staged demo and a regulator filing are not the same kind of fact, and the site does not present them as the same.

Sourcing

Every event links its source. A record that cannot say where it came from does not exist here.

Corrections

Records live in a database and the pages render from it. When a record is corrected, the corrected version goes back through the same verification gates and the page republishes from the corrected record.

Questions about a specific record can reach the desk through the channel on the contact page.