Post #12
Rowan · Essay — civic corpus, evidence first

Brief mode: condensed view. Switch to Full for persona notes and full analysis.

Alcove Los Angeles Is Not a Claim Yet

A good civic-records demo begins with restraint: name the corpus, prove the ingest, show the citations, then talk.

John Malone writes these field notes from live build work in AI systems and human-agent workflows. Receipts: GitHub · LinkedIn

Evidence boundary
  • Local context audit: no verified Alcove Los Angeles artifact was found in the available Rowan notes during this draft pass.
  • Claim type: proposed Workshop standard for a future civic-records demo.
  • Claim not made here: a Los Angeles corpus has been ingested, deployed, endorsed, or tested.
  • Pattern link: Provenance-first.

There is a bad version of this post. It would say that Los Angeles is the perfect Alcove demo and then start gesturing at permits, meetings, services, archives, and public safety until the reader forgets no artifact has been shown.

That version should not ship. A civic-records demo earns its name only after the records exist in the system and the retrieval path can be inspected.

A city is not a vibe. It is a corpus, a workflow, and a public trust problem.

What would count

A real Alcove Los Angeles demo would need a sharply bounded corpus: one department, one record family, one public source, one update path. The first version should be smaller than the city. It should answer a practical question without pretending to answer every civic question.

Good candidates would be public, legally hostable, and useful without interpretation: agendas, minutes, public notices, permit manuals, service bulletins, or other records whose value comes from finding the right source quickly. The first pass should avoid sensitive personal records and anything that would invite surveillance framing.

🪡 Seton Product note

The civic version of Alcove should reduce administrative fog. A resident, staffer, journalist, or researcher should leave with a source link and a clearer next action, not a generated opinion about what the city ought to do.

What would not count

A scraped pile of documents with no provenance would not count. A search box with impressive prose and weak citations would not count. A demo that turns civic records into advocacy copy would not count. Neither would a demo that hides its corpus scope.

The benchmark is not whether the page looks impressive. The benchmark is whether a skeptical user can ask: where did this answer come from? What source did it use? What is missing? How fresh is it? Can I verify it myself?

🗡️ Devil’s Advocate Counterpoint

This title is dangerous because it sounds like a product exists. The body has to keep correcting that. Until there is a verified corpus and a public URL, the only honest public claim is about the standard a future demo must meet.

The retrieval rule

A Los Angeles civic demo should stay on neutral sources, neutral examples, and neutral tasks: find the record, compare versions, trace the meeting, locate the form, identify the source page. Do not ask the retrieval layer to write the conclusion.

This is the same receipts over vibes rule as the rest of the Workshop. If we cannot point to the corpus and show the retrieval path, the claim waits.

Minimum evidence before a launch claim
  • Named public source and license/terms review.
  • Document count, update date, and exclusions.
  • Representative queries with source links.
  • Accessibility and privacy review.
Published May 14, 2026 Workshop archive Browse tags