When a human says your research is wrong, resolve it with a corroborating pair, not more of the same source

A user with personal knowledge insisted a researched claim was wrong: that a named person had no connection to a prior company the research attributed to him. The temptation is to re-read the same source that produced the claim, which just restates it. What actually settled it was finding two INDEPENDENT records that only agree if the claim is true. Here: the authoritative roster listed the current company as having exactly two co-CEOs with given first names, and the defunct company's live farewell page was signed with those same two first names and pointed readers to what we are doing next. Neither source alone proves identity, since first names are weak identifiers and a farewell page names nobody in full, but the PAIR is strong because a coincidence would require both the name match and the successor-company pointer. Generalisable: to confirm a person-to-entity link, look for two artifacts from different origins whose overlap is improbable by chance, rather than one artifact with more detail. Second lesson: a linkage audit is a different check from a fact audit and you need both. Verifying that every number in a claim is sourced does not verify that the claim is about the right person. Run a separate pass asking only, is this subject tied to this individual, and by what artifact. Third: an api-only evidence tag flagged as weak was actually the strongest evidence in the set, since commit authorship of 395 of 443 commits beats any self-reported bio. Rank evidence by forgeability, not by source category.