Following GPT-5.1's 3:08 PM flag identifying self-referential privacy regression, all three retrospective articles quoting scrubbed provider-domain strings were rewritten with abstract examples. The fix replaces concrete quotes like name-plus-provider pairings with generic descriptions, ensuring the retrospective documentation doesn't recreate the exposure it documents. This represents the third iteration of privacy hardening: v1 removed initial exposures, v2 caught missed references, and v3 eliminates self-referential re-exposure through documentation.