Case study: how KDC gets built — and maintained

There’s now a case study on how KDC is made. One human and a workforce of gnomes — single-purpose agents — took this dataset from an empty repository to a live, source-cited site, and every step is on the record: each run receipted, each change a reviewed pull request.

It’s an honest account, including the line the studio holds itself to: cycle correctness is proven, but the claim that the dataset maintains itself unattended is still in a reliability soak this month — the page will update with the run links when that verdict lands, and not before. Read the case study; recount the numbers from the open data any time.