The raw numbers behind the measurements, as JSON — including the control runs. Licensed CC BY 4.0, so they can be reused with attribution.
Whether a synthetic input event at window-system level satisfies the activeTab gesture requirement: tabs visible to the extension before and after a keyboard command sent through X11 XTEST. Method and discussion.
The counterpart: an agent driving the browser through CDP and the DOM only. The extension loads and its service worker runs, then every tab comes back with an empty url. Method and discussion.
For 150 cited web sources: live status, verdict, whether the Internet Archive holds a snapshot and how old it is. Cases where a service did not answer are marked as such rather than counted as negative. Method and discussion.
How much text survives when a full-page screenshot PDF goes through OCR, plus the full resolution series and a control run against unrelated text. Method and discussion.
Declared permissions, host permissions, user counts and ratings for eight PDF and capture extensions, as retrieved from the Mozilla add-ons API. Method and discussion.
Twenty sources of a reading list through a citation endpoint, each with the fields returned and — for every source handed back — the response to a reader and to a browser, so a page's wall can be told from a reader's limit. Method and discussion.
Eighteen scholarly platforms, each checked for the citation data it declares about its own works, with every returned title verified against Crossref. Method and discussion.
A mixed reading list of twelve sources: what an agent turned into citable records on its own, and the reason each of the remaining four was handed back. Method and discussion.
248 page-saving extensions from the Mozilla add-ons API with their declared Android support, user counts and ratings. Method and discussion.
Citation extraction measured against Citoid, the Wikimedia service built on the Zotero translators, across 18 works drawn at random — per source, with timings and every reported block re-checked from an unrelated network. Method and discussion.
The same article through the browser's print export and through full-page capture: page count, broken sentences, file size, recovered text. Method and discussion.