A small entry about a small mistake, and what it took to actually stop making it.
During the drafting of Article 5, something in the closing section read wrong. Not wrong in the sense of a typo or an awkward sentence. Wrong in the sense of being oddly familiar.
The final line of the article, the one built to stand alone as a pull quote, was word for word the same sentence already used as the Key Takeaway a few sections earlier. Same words, same order, just appearing twice, once early, once at the very end, dressed up in two different visual treatments as if they were two different ideas.
That was the second time it had happened. The first was Article 4, where the same thing was caught and fixed before publishing. Seeing it again on Article 5 wasn’t a new problem. It was the same problem, missed the same way, twice.
The fix itself was almost anticlimactic. Nothing needed to be invented. The article already contained a different sentence, one that had been written earlier in the piece and never used as a closer, that could stand on its own just as well. The duplicate got replaced with something that was already there. No new writing, just noticing what existed.
That would have been a tidy enough story on its own. But during a later review pass, the same article turned up a second duplication, a different repeated line this time. The first correction hadn’t ended the review. The article was looked at again, and something else was found.
What had actually changed wasn’t the proofreading. Every sentence in both drafts read fine on its own. What changed was the question being asked. Not “is this sentence good,” but “does this sentence quietly repeat something the reader already saw, two sections and a different visual box ago.” That’s a harder thing to catch, because nothing about either sentence, read in isolation, is wrong.
That shift, checking whether the finished page actually held together, not just whether each part of it was well written, became one of several lessons that shaped how every article since has been checked against the live site rather than assumed correct from the draft. Other problems later reinforced the same lesson from different angles: a corrupted image that passed structural checks but wasn’t actually right, and a formatting bug that only became visible once the text was live.
There’s no clever process behind any of this. The fix that held wasn’t a system. It was noticing something that had already been written.