Public data is published. It is not served.

Enormous amounts of useful information are legally public and practically unusable. A federal agency posts a file. The file is large, oddly shaped, updated on its own schedule, and documented in a PDF. Anyone who wants to build on it spends weeks writing a pipeline before writing a single line of the thing they actually wanted to make. Then they maintain that pipeline forever.

Vemon does that part once, properly, and serves the result through an API. The data is not ours and never will be — the value is in collecting it reliably, normalising it honestly, versioning it so you can reproduce a result, and keeping it up.

Where we are

One dataset, 63,518 records, covering 49 states and DC. That is a deliberately narrow start — one source, done properly. A catalog of a hundred half-maintained datasets is worth less than one that is correct, documented, and reliable, and the second dataset is much cheaper than the first once the pipeline is real.

How we intend to behave

Numbers on this site are computed, not written. Every count and statistic comes from the dataset at build time. If a figure looks small, it is because it is.

Provenance is published. Every dataset page names its source, its data year, and what we changed during ingestion. The ingestion script is public.

Caveats sit next to the data, not in a footnote. Medicare submitted charges are not what patients pay, and the healthcare dataset says so on the page where you query it — not because it is legally required, but because a number without its caveat is worse than no number.

Unbuilt things are labelled.Anything marked “soon” on this site does not exist yet.