Scan log and change record

What the automated job does

Once a day a scheduled workflow scans public news sources, writes any candidates to a review queue, rebuilds the site from the existing ledger, commits generated files and uploads the static output. Candidates do not go straight into the ranking. A row only becomes a published event after the de-duplication rule runs, and an unreviewed row keeps its ⚡ flag and source URL until an editor checks it.

The distinction matters because a scan that finds two hundred articles is not a day with two hundred new financings. Most are unrelated, syndicated or about programmes rather than priced rounds. The counts below separate those stages rather than presenting one impressive total.

Editorial rule changes on this site

The publication layer now merges syndicated reports within a ten-day window. Generated interpretive sentences were removed entirely; every published event carries its own hand-written note, and an event without one renders factual labels only. Ranking subpages gained individually written analysis instead of a repeated title suffix. Company profiles are generated only for companies with at least one published financing, because an empty profile would be a doorway page rather than a useful document.

Each of those decisions changed what a crawler sees but not what the raw ledger contains. The corrections log records fixes to individual figures; this page records changes to the counting and rendering rules.

How to read the numbers below

Raw reports is the size of the source ledger. Published events is what remains after merging. Pending review is the number of candidates found by the last scan that have not yet been checked. The data cutoff is the newest human-reviewed event in the dataset, and is deliberately shown separately from the date this page was generated, so a build timestamp is never mistaken for fresh evidence.

Current scan and publication state

Ten most recent published events

The practical takeaway

If you are evaluating whether this site is maintained, the two dates and the review-queue size on this page are more informative than the word daily in a page title. We publish them precisely so that claim can be checked.