I wrote 32 checklists, then went back for the numbers I could not prove
This site launched with thirty-two generators and one uncomfortable document: a file called DATA-SOURCES.md whose first line reads NOT YET VERIFIED AGAINST PRIMARY SOURCES.
That deserves an explanation, because it is not how the sibling project was built.
What happened
The checklists were written from general trade and agency guidance — the kind of thing that is consistent across the field and that anyone who has worked around houses would recognise. Filter intervals, septic pumping frequencies, grading falls, water-per-person-per-day, the ages at which water heaters fail.
None of it was transcribed from a document that was open at the time.
For most of it, that is fine. “Six inches of fall over the first ten feet” is not contested. “Clean the gutters after the leaves are down” needs no citation. But a checklist is not made of general guidance — it is made of specific numbers, and some of those numbers are about law, money or safety.
Those are different. Being confidently wrong about when to clean a gutter costs nobody anything. Being confidently wrong about how long you have to change a driver’s licence, or what a policy caps jewellery at, or how long a carbon monoxide sensor lasts, costs a reader something real.
The one that was probably actually wrong
Going back through it, the item that stood out was the alarms.
Several pages said, in effect: test every smoke and carbon monoxide alarm, and replace any unit over ten years old.
Ten years is the figure people know for smoke alarms. Carbon monoxide alarms use a different sensor with a different service life, and they are frequently rated for fewer years — which means a sentence treating them as one category is quietly telling some readers to keep a dead CO sensor for years after it stopped detecting anything.
The fix was not to substitute a different number, because I had not verified that one either. It was to stop asserting a shared figure at all:
Every unit has a manufacture date on the back and an expiry stated in its own manual — smoke alarms are commonly rated ten years and carbon monoxide sensors often less, so check both rather than assuming they age together.
That sentence is longer, less satisfying, and correct.
Four more that got softened
Same treatment, same reasoning — each was a specific number about law or money that had not been checked:
- The moving damage claim window. The page named a figure. It now says the deadline is in your paperwork, that it is shorter than you expect, and that it starts on delivery.
- The jewellery sub-limit. It named a dollar range. It now says the cap is far below what a household owns, that it is a total rather than per item, and to read the number off your own declarations page — which is more useful advice anyway.
- The flood insurance waiting period and the Closing Disclosure three-day rule both stayed, because both are well-established rules rather than rules of thumb. Both are on the verification list, at the top.
Why publish the list instead of quietly fixing it
Because the pages themselves make an argument about exactly this.
The HVAC page tells readers that filter manufacturers publish shorter intervals than technicians recommend, and that this is a fact about who is selling filters. The maintenance page says that where a source disagrees with your equipment manual, the manual wins. The whole editorial position is be suspicious of confident intervals from people with an incentive.
A site that says that, and then presents its own unverified figures with the same confidence it just warned you about, has not understood its own argument.
What happens next
docs/DATA-SOURCES.md lists every figure on the site, what it is currently based on, and the specific source to check it against. It is ordered by consequence: the things that are numbers about law, money or safety first, and the things where being roughly right is fine last.
The pages will not change much. Most of the figures are the common ones and will check out. But “will check out” and “checked” are not the same claim, and until the second one is true, the file says so.
If you find one before I do — a number that contradicts your own equipment manual, your state’s rules, or your policy — that is the most useful message you can send.