Cream City Almanac

EACH WEEK

1. Read DYCU's Hub catalog and its inventory sheet.
2. Group the yearly and geographic versions of each measure into one family with a permanent code, like F02.
3. Read the 180 report PDFs so search can match the text inside them.
4. Ask AI (Claude) to write each family's plain-English explainer, column guide, caveats and story angles. DYCU's own definitions always win.
5. Index everything for search by meaning and by exact words.

A failed week never replaces the live catalog. Data as of Oct 8.

WHAT THE AI DOES

Every fact carries a tag saying who wrote it:

• DYCU DYCU's own definition, from its inventory sheet.
• HUB From DYCU's Hub listing.
• SOURCE From the source agency's own website.
• AI Written by AI from the facts above, and always labeled.

Numbers on a sheet come from the Hub's data, never from the AI.

HOW SEARCH WORKS

Type what you're reporting on, in your own words. Search matches meaning (text that means something similar) and exact words, then ranks what agrees. Try kids who can't afford food: it finds Food Insecurity Prevalence.

When nothing is close, it says so instead of listing unrelated data.

WHAT IT IS

An unofficial, phone-first way to find, understand and download Milwaukee's public data from Data You Can Use. Search in your own words; every fact shows where it came from.

WHERE THE DATA COMES FROM

DYCU's ArcGIS Hub, the public website where Data You Can Use publishes its data: 93 raw data · 282 reports · 7 visualizations, in the Hub's own words.

Refreshed weekly and never altered. Downloads come straight from the Hub.

HOW IT'S BUILT

Next.js on Vercel; Convex for the database, search and the weekly job; Claude through the Vercel AI Gateway; Firecrawl to read PDFs. A weekly automated check grades search on 26 reporter-style questions. Every trade-off is in the decision log.