People who work in City government
A general walkthrough for people inside the thing being queried. Two chapters go deep: what the open data portals do that a search engine structurally cannot, and what Checkbook shows you about a vendor that your own agency's system never will.
What the session covers
Every other script in this repo demos to someone outside government looking in. This audience works inside it, which inverts two assumptions the other scripts make: they already have better data than you about their own operations, and “accountability” is the wrong frame. The same Checkbook query a watchdog hears as oversight, this room hears as procurement intelligence it is currently missing.
- The open data portals, in depth. Five slides on why a search engine cannot answer a question like “how many 311 complaints has Brooklyn filed since July 1, by type.” A search index returns documents somebody already wrote. An aggregate over live rows is not a document.
- Checkbook NYC, and the cross-agency view no single agency can assemble: one vendor, every agency, one fiscal year.
- The other seven connectors, at a glance rather than in depth.
- What happened when we reported a problem to an agency — answered in two days, and the residual fault turned out to be ours.
- How BetaNYC runs on the same tools: the six-phase newsletter pipeline and its human gate, a WCAG 2.1 AA audit of our own site, and a membership-gated Discord that revokes access nightly.
The finding the Checkbook chapter is built on
Asking Checkbook what the City paid Microsoft in FY2026 returns 49 checks and $14,482,518.24 across roughly two dozen agencies. Complete, correct, verifiable, and still not the number you wanted.
The City’s Microsoft enterprise license agreement is $53,006,444.85, in two checks, paid to Dell Marketing LP. Most large software is bought through a reseller, so the vendor on the check is not the vendor you asked about, and a payee-name search never sees it.
Do not sum those two figures. We have not established they are disjoint, and the claim on the slide is about the search, not about a citywide total. The deck says so on the slide and again in its presenter notes.
What it is honest about
The deck concedes on its second slide that a published dataset is a poor substitute for an agency’s own system of record, and it ends by admitting we cannot tell anyone whether their IT policy permits installing any of this. Both concessions are load-bearing with this audience.
The Checkbook and Socrata figures were run live on 2026-08-04; three others are carried from earlier work and keep their own dates. Every number in the script’s provenance table names the call behind it. Where our own documentation contradicts itself — the newsletter scanner is “~130 sources” in two files and “~115” in a third — the presenter notes say so rather than defending a figure we cannot source.