The thesis
Local law governs daily life — and most of it is dark.
The rules that decide whether you can keep chickens, run a business from home, or build an ADU are written by 22,628 local governments. Far too little of that law is free, machine-readable, or comparable across places. Atlas Civica starts from the full Census universe and treats every gap as a target — lighting up the map as the engine frees more.
How the engine works
Atlas Civica is built entirely on Cloudflare. A harvester pulls codified law from seven codification vendors — Municode, eCode360, American Legal, Code Publishing, Sterling, QCode, and Franklin — preferring structured/HTML/text sources over OCR. Every section is mapped to the Atlas Civica Format: a stable URN, an FRBR work/expression identity, and a content hash.
A freshness gate avoids re-pulling unchanged codes; a diff engine turns every content-hash change into a dated, queryable version. An ML layer adds function and normative scores, a 42-subject taxonomy, and semantic vectors. The result is served as two metered APIs, an analytics cube, and this public atlas.
For researchers
The corpus & labels
The data is independently sourced — harvested directly from the codification vendors, not derived from any third-party dataset. Every labeled section carries a function class, normative scores, and subject tags, so findings stay comparable across the corpus.
| Capability | Atlas Civica |
|---|---|
| Corpus breadth | 9,700+ vendor targets, harvested continuously |
| Source | Direct structured / HTML / text — no OCR |
| Function taxonomy | 5 classes: Context, Rules, Process, Enforcement, Structural |
| Normative dimensions | Opacity, paternalism, enforcement discretion, strictness |
| Subject taxonomy | 8 domains / 42 leaf subjects |
| Semantic search | Vectorize bge-base + /v1/search |
| Identifiers | GEOID + OCD-ID + URN + FRBR work/expression |
| Liveness | Continuous re-crawl + content-hash diffs + dated versions |
Legal posture
Why we can republish the law
Under Georgia v. Public.Resource.Org (2020) and the government edicts doctrine, the binding text of the law is not copyrightable — no one holds a copyright in the law a government enacts. Atlas Civica republishes only the enacted legal text, with provenance preserved per provision (source_url, vendor, pulled_at). We do not redistribute vendor-proprietary annotations, editorial apparatus, or page-layout assets.