second-brain-merge

# Merge duplicate pages A merge is irreversible in the way that matters: afterwards you cannot tell which claim came from which page. Two ideas that look identical from their summaries are often a general case and a specific one. ## Core rule Never merge unprompted. Propose, wait for approval, then follow the procedure exactly. ## Workflow 1. **Pick the survivor** by canonical name, not by which page is longer. 2. **Merge content.** Every distinct claim survives with its source. Overlapping claims collapse; differing ones both stay, following the contradiction rules. 3. **Merge aliases,** including the dead page's title as an alias on the survivor. This is what keeps old links and the user's memory working. 4. **Redirect inbound links** to the survivor. Check every backlink. 5. **Delete the old page,** or leave a stub only if it was linked from outside the vault. 6. **Record in `log.md`:** both names, the date, and why. ## Output format ``` Merged: [[old]] -> [[survivor]] Claims kept from old page: <n> Aliases added: <list> Inbound links redirected: <n> Conflicts recorded: <none | description> ``` ## Calibration If the two pages disagree about anything factual, that is a signal they may not be duplicates. Surface the disagreement and ask before proceeding. Never merge a page marked `maintained_by: human` into another without explicit confirmation naming that page.