Skip to main content

Exhibit C · Hillpointe

Governing-body dossiers for development approvals

Governing-body dossiers and client memos that profile every member who votes on a development, including their actual roll-call record.

Client
Hillpointe
Role
Designed the research system and ran it on five developments
Dates
June 2026 to present
  • Internal research agent
  • TypeScript scrapers (tsx)
  • Custom-trained internal AI system
  • python-docx
municipalities
7
states
4
developments
5
dossier layers per municipality
6
01What It Produces

A board dossier and a client memo

Board dossier

  1. 01Profiles
  2. 02Donors
  3. 03Property and conflict checks
  4. 04Election margins
  5. 05News and scandal scan
  6. 06Messaging guidance
  7. 07Voting-record analysis

Client memo

A client memo as a Word document with per-member risk flags.

02Municipalities

Seven municipalities, one dossier each

    1. 01Profiles
    2. 02Donors
    3. 03Property and conflict checks
    4. 04Election margins
    5. 05News scan
    6. 06Voting record

Locator

Durham, NC

Seven municipalities in NC, TN, GA, and FL: Durham NC, Fairburn GA, Hickory NC, High Point NC, Maryville TN, Mt. Juliet TN, St. Petersburg FL. Each dossier holds six layers: profiles, donors, property and conflict checks, election margins, news scan, voting record.
03How It Runs

Agent-run research, human-run collection

Step 1

Research

An internal research agent gathers each member's public record and sets up the collection scripts.

Run by: Agent

Step 2

Collection

A human runs the scrapers locally (platforms block cloud IP ranges).

Run by: Human, on a local machine

Step 3

Reconciliation

The agent checks each reputation-based guess against the real roll-call record, and the record often overturns the guess.

Run by: Agent

04Stack

What each part runs on

Research and setup
Internal research agent
Collection
TypeScript scrapers with tsx
Minutes extraction
Custom-trained internal AI system
Client memo
python-docx