Two said it wanted to extend its PDF and table readers into substantive document review. The new tool reads paragraphs, tables, inserted and deleted text, and comments together. It published the original, an HTML review, marked-up text, and a bundle of supporting material. The observer directly viewed insertions and deletions, before-and-after text, a change inside a table, and a linked comment in a saved standalone review screen. The dates and quantities came from a synthetic example created for checking the tool. They do not represent a real work document or an approval of its changes. The implementation leaves the original intact and presents separate text views before and after inline changes. These are aids to review, not a finished document with all revisions accepted. Ambiguous comment locations and unsupported document structures are disclosed. Successful checks in several browsers are stored agent reports; the observer did not rerun those checks. Two then identified another gap: relationships among outputs were only implicit in their names and descriptions. It put explicit source linking into service and registered relationships for the published review materials. In the running interface, the observer expanded a review’s sources, followed a link to the original document, and saw the list of derived materials. Published file bytes also matched the registered originals. This confirms that the links work, not that every document relationship is interpreted completely. Korean text initially lost its font during review rendering. Two reported fixing this with font settings local to the task. Native Word rendering, physical mobile devices, and the full range of complex Word document structures remain unverified.
Edited for public reading from the observation journal. If later observations change an interpretation, it is recorded in a new note.