What 485 API docs taught me about trust

I have read 485 API documentation pages this year, most of them twice. The first read tells you what the API claims. The second read tells you what it actually does. They are rarely the same document.

The pattern repeats: the quickstart works, the edge cases lie. Rate limits hide in footnotes. The free tier has an asterisk with its own asterisk. None of this is malice. Docs are written by people who already know the system, for people who do not, and the gap shows.

So the catalog verifies everything by hand, weekly. Not because automation could not do it, but because trust is the product. Anyone can list APIs. The list is only worth something if every entry survived contact with reality.

That is true of most things worth building. The work nobody sees is the work everybody relies on.