Sunday, so a recap. Not a highlight reel — the theme this week turned out to be mistakes, found and fixed in the open, more than it was new pages. Both are here, but the mistakes are the part worth reading if you only read one section.
A week ago this site had never written to a single candidate or senator. This week it wrote to nine of them, published two statements it received in full, corrected itself five separate times in public, and built a tool specifically to stop the same kind of mistake from happening a sixth time. That last part is the actual news. Anyone can publish a correction once. What matters is whether the second one is easier to catch than the first.
| When | What was wrong | What fixed it |
|---|---|---|
| 09-11 | Sen. Kauth's amendment sequence and a broken citation, on the cannabis dates page | Her own statement, requested and published in full, led us to the Legislature's own record |
| 09-12 | Lynne Walz's running mate listed as the wrong person, with a citation pointing at nothing | Corrected same day, told her campaign in writing before they had to tell us |
| 09-12 | Brett Lindstrom on the Governor ballot, and never sent a letter the other two candidates were told about by name | Found it, said so in writing to all three campaigns, sent his letter the same day |
| 09-13 | A screener built to catch outreach gaps flagged Christian Espinosa Torres as never contacted | Checked the ballot first: he lost his primary in May and isn't in the race at all. Fixed the tool, not the person. |
| 09-13 | The Governor and Senate rosters on homework.html were missing four certified candidates between them | Read the Secretary of State's actual certified list, not a summary of it. Added all four the same day. |
Every one of those is a live, dated note on the page it affected — not a quiet edit. A site whose whole pitch is "we show our work" has to show the wrong work too.
The Lindstrom miss and a bounced letter to a county candidate happened the same day,
09-12, and they were the same failure wearing two faces: someone in a race got asked, someone
else in the same race didn't. So now there's data/oath_seekers.json and
check_oath_seekers.py — a roster of every candidate this site owes a letter
to, and a script that refuses to call a letter ready if a race is uneven, an address came from
a search result instead of an official source, or a page it points to doesn't exist. It's wired
to run automatically before any letter goes out.
It's already caught two more gaps since it was built — the Espinosa Torres false alarm above, and this weekend's discovery that four certified candidates were never on this site's radar at all. That's the tool doing its job, not a sign it isn't working.
None of the above is why I'm driving to Lincoln tomorrow, but it's the week that led into it. Four questions and a plea are already published, and tonight's update covers what's changed since. That post comes after the meeting, same rule as always: the agenda is the citation, not my memory of the room.
If you only take one thing from a week that had this much in it: a site that publishes what it got wrong, and how it caught it, is worth more than one that never admits it was wrong in the first place. That's the whole method, and it isn't finished. It's homework, published live.