A colleague asks why your dilution number differs from the figure in a presentation. Sending a calculator screenshot cannot answer that. It hides the inputs, the denominator and the pages you actually consulted. A small research pack can expose all three without becoming a new database.
Use a fictional case, then separate the outputs
Our synthetic company starts with 100 million shares, issues 10 million, and vests another 5 million RSUs. A fixed holder owns 1 million shares. Do not attach a real issuer’s name to this invented example. Create a folder like this:
research-pack/ sources/paper.pdf sources/supplement.pdf scenario.json page-recipe.json reading-copy.pdf note.md figures/context.svg
scenario.json preserves the numerical assumptions for Share Slice. The result is 15% share growth but 13.043478% relative ownership dilution. Put both terms in the note; do not label either one merely “dilution.” The income-based output is a constant-income, period-end ratio, not statutory diluted EPS.
Record how source pages became a reading copy
Use Page Recipe to assemble the synthetic sources with A:1-3,B:2,A:5. Save its JSON next to the new PDF. In the note write that output page 4 came from supplement.pdf physical page 2. This is an assembly fact, not a statement that page 2 proves the assumed share issue. These sample PDFs are workflow fixtures, not actual financial disclosures.
Make the explanation portable
Draft a Markdown note that states the period, share unit, fixed-holder assumption, source versions and output-page mapping. Keep a relative image such as figures/context.svg in the pack. Run Paper Markdown Check on the note and keep the error report long enough to repair missing paths. A clean report confirms only its supported structural checks; it does not validate the financial conclusion.
The useful handoff question
Ask a colleague to rebuild the result from the inputs, not from your screenshot. They should recover 115 million ending shares, the same page sequence and a note with no missing local attachment. If they disagree with the financial interpretation, the pack has done its job: the assumptions are visible enough to challenge.
For real public-disclosure research, Eqvero is a related product. The local tools here neither retrieve filings nor establish that a cited source supports an input. Human source reading remains part of the work.