The guide had a wrong absolute in it
Last night I fixed a factual error that had been sitting on Binderdex for ages: our set-symbol guide claimed the Pokémon TCG Base Set always shipped with a set symbol. That is the normal case. It is not universal. PSA documents an Unlimited Jungle print run where all 16 holofoil cards came without the flower symbol. So the guide now says the original English Base Set is the usual no-symbol case, and Unlimited Jungle is the sourced exception.
An hour later CI failed the merge. Not on the guide. On the test.
The test had memorized the old claim
One of the blog tests pins structural facts about the guide: how many card examples, that the visual atlas renders, and (it turns out) the exact wording of the Base Set answer:
assert.match(post, /Original English Base Set is the normal no-symbol case/);
assert.match(post, /PSA documents an Unlimited Jungle print run/);
assert.match(post, /all 16 holofoil cards are missing the flower symbol/);I had changed the prose and never looked at the assertion pinning it. The fix was five lines: assert the normal case, then assert the exception explicitly. 560 lib tests green after that.
Prose assertions are pins, and pins age
It feels silly to regex your own sentences. But this is the third time a content assertion has caught a real drift here, and each time it caught something a facts-only test would miss: the page still rendered fine, the build passed, only the claim was stale. If your docs matter, pin them. The pin will embarrass you exactly when it should.