Module 1 · ~10 min

🧭 Find an idea & vet the data

You'll leave with a real dataset you've confirmed exists — and one sentence describing the app it becomes.

The idea is downstream of the data

Most "app ideas" die because the data behind them was imaginary. Flip the order: start from a dataset that already exists and is public, and the idea falls out of it. Every app in the fleet started as "here's real data nobody made legible yet," not "wouldn't it be cool if."

Four patterns that reliably turn a dataset into an app:

• Searchable — a public records set with no search box. Add one.
• Mappable — anything with an address or lat/lng wants to be a map.
• Personal — "what does this mean for my street, my bill, my zip?"
• Legible — a raw CSV turned into a chart a human actually understands.

Where the good free data hides

• City/county/state open-data portals ("<place> open data").
• ArcGIS REST services — assessors & GIS departments publish queryable layers (parcels, owners, zoning). A goldmine most people never touch.
• US Census geocoder — turn addresses into coordinates. Free, no key.
• open-meteo — weather, air quality, marine. Free, no key.
• Agency feeds — transit (GTFS), rain gauges, air sensors, court records, FOIA drops.

The 20-minute vet — do this BEFORE any code

1. Find the source, not the topic. "Phoenix evictions" is a topic. A specific endpoint that returns JSON rows is a source. If you can't name the endpoint, you don't have a project yet.
2. Open it raw. Paste the API / CSV / ArcGIS URL into your browser. You should see real rows within minutes. If you can't, the app is telling you no.
3. Check the three killers (one strike is usually fatal):
   • Complete? — or riddled with nulls, only three counties, "coming soon."
   • Fresh? — updated this month, or last touched in 2019?
   • Allowed? — do the license/terms let you use and display it?
4. Check the shape. How many rows? Does it carry the field you'd build around (a location, a price, a date, an id)? Can you join it to something (address → coordinates)?
5. Decide. Green on all three killers → build. Any red → change the idea or the source. Killing it now is free; killing it after you've built is not.

Let Claude Code do the probing

You don't have to eyeball raw JSON yourself. Paste this into your editor and let it report back:

I'm vetting a dataset before building anything. Endpoint: <PASTE URL>

Fetch it and tell me, concretely:
1. Does it return usable data right now? Show me 2 sample records.
2. Coverage/row count — complete, or partial?
3. Freshness — is there a date field, and what's the newest value?
4. Fields I could build around (location, price, date, id)?
5. Any license, terms, or rate limit I should know about?
6. Verdict: worth building on? If not, why not?

Be blunt. I'd rather kill this now than after I've built it.

Your turn ✅

• Name one dataset you can open raw in your browser right now.
• Run it past the three killers — complete, fresh, allowed.
• Write one sentence: "This makes ___ data searchable / mappable / personal for ___."
• If any killer is red, swap the source before you move on.