ResearchWorkflow
10-K / 10-Q Diff
Last updated April 27, 2026
What it does
Drop two filings into a folder. Cowork tells you what materially changed in the risk factors, MD&A, and accounting footnotes — quoting the actual language, summarizing why it matters, and flagging the items that warrant a closer read. Structural changes in disclosure language are alpha-relevant and almost never read in full.
What you need first
- ·Two filings in a folder, named clearly:
aapl-10q-q1-2026.pdfandaapl-10q-q4-2025.pdf - ·A working knowledge of the sections that matter to you (most analysts care about Risk Factors, MD&A, Revenue Recognition footnote, Segment footnote, Subsequent Events)
The prompt
Compare the two filings in this folder for {COMPANY}.
For each of the following sections, identify what materially changed between the two filings:
1. Risk Factors — new risks added, old risks removed, language sharpened or softened
2. MD&A — narrative shifts on growth drivers, headwinds, and forward outlook
3. Revenue Recognition footnote — any change in policy, timing, or judgments
4. Segment reporting — segment-level shifts, any reclassifications
5. Subsequent Events — anything noted post-period
For each material change:
- Quote the prior language and the new language side by side
- Write 1–2 sentences on why it matters
- Rate the materiality: HIGH / MEDIUM / LOW
Save as a markdown file named "{TICKER}-10K-diff-{date}.md" in this folder.
If the language change is purely stylistic and not substantive, skip it. I want material differences only.What good output looks like
- ✓Side-by-side quotes, not paraphrases
- ✓Rated materiality so you can scan
- ✓Stylistic-only changes correctly excluded
- ✓Reasoning attached to the “why it matters” — not generic
Make it a Skill
Save as 10k-diff. Run with two filenames as arguments and the materiality threshold you want. Useful as a Skill because the structure should look the same every time so your team can scan diffs across the portfolio in a consistent format.
Common pitfalls
- Big filings, big context. A 200-page 10-K is a lot. If output is incomplete, run section-by-section and combine — see the drift trap chapter.
- Materiality is judgment. Calibrate the first run with your PM's eye — what “HIGH” means to you needs to be encoded into the Skill instructions.