An interactive grid for exploring institutional design for AI governance. Maps institutional mechanisms (protocols, preferences, rights, incentives, expertise, norms, thick commitments) across scales (dyadic, group, community, national, global).
Three perspectives:
- AGI Institutions — new institutions needed for a world of autonomous AI agents
- Existing Human Institutions — current institutional infrastructure and how humans accomplish alignment
- Fidelity & Meaning — institutions to align organizations with rich, accountable mandates
Each grid cell expands to show detailed frameworks, and many include problem sets designed for pairs or small teams (~1 hour each).
Content lives in data/{agi,human,fidelity}/ as markdown files named {row}-{col}.md (e.g. dyadic-protocols.md).
Each file follows this format:
# Cell Title
Body content rendered in the detail view.
## Researchers
- [Name](url) — brief description of their work
## Key Papers
- [Paper title](url) (Year). Summary of relevance.
## Learning Resources
- [Resource title](url) — textbook, tutorial, course, etc.
## Problem Sets
### Problem 1: Name
Problem description and deliverables.Cells can include any combination of these sections — all are optional. Use whatever standard markdown structure fits the content.
To add a new cell, create the markdown file and run npm run build. The grid links also point to GitHub for easy editing.
npm install
npm run db:migrate:admin-auth
npm run buildThe app uses React Router SSR. Copy .env.example to .env and provide the database
connection plus Mailgun/session settings before testing the authenticated
/researchers/admin routes.
The public site remains open. /researchers/admin is protected by passwordless email codes sent
through Mailgun to addresses already present in researchers.email. Codes expire after
10 minutes and successful verification creates an HttpOnly session cookie.
Hosted on Vercel. Push to main to deploy; Vercel runs the React Router build. Configure
the variables in .env.example in the Vercel project.