Skip to the content.

A one-hour hybrid seminar on the paper. A short presentation on agentic AI in social-science research, threaded with three live demos you can code along with. Two principles anchor the argument: researcher control and radical transparency.

⬇ Download the workshop package (ZIP) ⚙ Install Claude Code

What you’ll walk away with

Setup

After installing Claude Code, authenticate with your Anthropic account - trial credits cover the session. Unzip the workshop package wherever you keep research projects, then open workshop-package/ itself in Claude Code or Codex so the local wiring loads.

Before the session, create exactly one working folder under workshop-package/projects/:

The package ships with local CLAUDE.md, AGENTS.md, and Claude-native wrapper skills, so it does not depend on a separate personal AI Operating System install.

Data: bring your own, or use ours

Option A - bring your own. Pick one concept from your current research that you have struggled to measure directly - something latent. Examples: civic engagement, clientelism, state capacity, democratic backsliding, institutional trust, political efficacy. Bring:

Option B - use one of our toy projects. The package ships three; pick the one closest to your field.

Two questions to think about beforehand

  1. What is one concept in your research you have struggled to measure?
  2. What validity threat have you worried about and not fully resolved?

Format

Sixty minutes, integrated Q&A - interrupt whenever rather than saving it for the end. Three live demos of about five minutes each. The slide deck and the package contents are yours to keep.

After the seminar

The workshop package is self-contained. If you want to keep building after the session — composing your own skills, layering personas, wiring multiple projects — the public scaffolding behind Simone’s personal AI OS is open-sourced as a starter library at github.com/simonepaciphd/ai-os-public. It ships the five personas (chief-of-staff, researcher, writer, engineer, teacher), protocol skills, and a librarian subagent. Methodology to fork and customize, not a turnkey product.


Questions before the session: spaci@stanford.edu