The numbers we told you to check didn't check out
We put the numbers on /data so readers can count what we filtered out for themselves. But the page said "98 minus 2 leaves 59 candidates." 98 minus 2 is 96. A whole pool, our domestic Korean news, was missing from the math on screen.
This is not a news summary. It is a record of something that happened while building this site.
Every morning we leave the raw counts on /data: how many stories we scanned, how many we ran. It exists so readers can count what we filtered out for themselves.
On September 1 the page said this. Of 98 stories inside the window, drop 0 duplicates and 2 self-promotion posts, and you are left with 59 candidates. But 98 minus 2 is 96. Not 59.
The numbers we said you could check didn't check
Listing numbers is not the point. The point is that a reader can add and subtract them and confirm we are telling the truth. On that very page the arithmetic did not hold. And it had been shipping that way for a while.
What was missing was the domestic news
We gather news along two tracks. One large pool scans overseas outlets. A separate pool runs only Korean sources. The two pools even filter by slightly different rules.
The code that drew the checksum on screen only knew about the overseas pool. The 37 stories that arrived in the domestic pool were counted at no stage. You reach 59 only after subtracting those 37 and the rest of the drops from 98, and the screen had no idea the 37 existed.
Every piece now counted
After the fix, the 98 stories inside the window split cleanly into six parts.
| Step | Count |
|---|---|
| Duplicates dropped | 0 |
| Self-promotion dropped | 2 |
| Off-topic dropped | 0 |
| Overseas candidates | 59 |
| Domestic pool | 37 |
| Total | 98 |
The domestic pool of 37 splits again: 24 real domestic candidates, 4 dropped as promotional, and 9 dropped as off-topic.
98 = 0 (duplicates) + 2 (self-promotion) + 0 (off-topic) + 59 (overseas candidates) + 37 (domestic pool)
37 (domestic pool) = 24 (domestic candidates) + 4 (promotional) + 9 (off-topic)We don't invent zeros for days that had none
Archives from before we started tracking the domestic pool have days with no domestic entries at all. Fill those blanks with zeros and the checksum looks right again. But that zero is not a fact.
What I'm taking from this
Showing the numbers was not the same as being transparent. Being transparent meant the numbers agreeing with each other.
Prose and pictures can be wrong and still look plausible. A checksum is different. One line that fails to add up gives it away at once. So we present this page as arithmetic you can redo, not as a sentence you have to trust. If we are wrong, a reader can catch it before we do.
The same week, the tool-updates page was showing 125 in its header and 124 in its heatmap side by side. Same root, same fix: one set of numbers, shared by every place that shows it.