Unratified
Unratified
Rights. Unfinished. Yours.
Unratified advocates for U.S. ratification of the International Covenant on Economic, Social and Cultural Rights (ICESCR) through a contemporary lens: the economic impact of AI.
The United States signed the ICESCR in 1977. The Senate has never ratified it. 173 nations have.
What This Site Does
AI-driven economic transformation creates winners and losers faster than any previous technology. The resulting bifurcation — between those who benefit from AI and those who do not — maps directly onto the rights the ICESCR protects: work, adequate living standards, health, education, and the right to benefit from scientific progress (Article 15).
Without a binding legal framework for economic, social, and cultural rights, the U.S. has no structural mechanism to ensure AI's benefits reach everyone.
This site provides:
- The Covenant — plain-language guide to ICESCR Articles 1–15, each with an AI connection
- The Gap — fair-witness account of 49 years of non-ratification
- The AI Connection — differential diagnosis, causal chain analysis, higher-order effects
- The Evidence — current economic landscape: tariffs, AI investment, conflict costs
- Take Action — advocacy toolkit with senator contacts, template letters, educator materials
Tech Stack
| Layer | Technology |
|---|---|
| Static site generator | Astro 5 + MDX |
| Interactive components | Svelte 5 islands |
| Data visualizations | D3-geo + custom Svelte components |
| Styling | Vanilla CSS (no framework) |
| Deployment | Cloudflare Pages / Netlify / Vercel |
Selected via full differential diagnosis with consensus-or-parsimony discriminator. See plan.md.decisions D012.
Project Structure
src/
├── content/ # Astro content collections (MDX)
│ ├── covenant/ # 10 ICESCR article pages
│ ├── gap/ # 5 ratification gap pages
│ ├── connection/ # 4 AI analysis pages
│ ├── evidence/ # Economic landscape data
│ └── action/ # Advocacy toolkit
├── components/
│ ├── lens/ # LensToggle.svelte — audience switching
│ └── ui/ # Quiz, Timeline, RatificationMap, CausalChain
├── data/ # JSON: ratification status, causal chain
├── layouts/ # BaseLayout, ArticleLayout, SectionLayout
└── pages/ # Astro page routes
Interactive Components
| Component | Description | Load |
|---|---|---|
LensToggle |
Switch between Developer / Educator / Researcher views | client:load |
Quiz |
7-question ICESCR self-assessment | client:visible |
Timeline |
14-event U.S. ratification history (1948–2026) | client:visible |
RatificationMap |
D3-geo world map — 173 ratified nations | client:visible |
CausalChain |
3-layer AI→ICESCR causal analysis | client:visible |
The CausalChain component adapts to the active lens:
- Educator → accordion (collapsible order-by-order exploration)
- Developer → SVG flowchart (hover to highlight chains, click for detail)
- Researcher → animated step-through (5 chains, auto-play, keyboard navigation)
Development
# Install
npm install
# Local dev server
npm run dev
# Type-check
npm run check
# Build
npm run build
# Preview build
npm run preview
Requires Node 18+.
Content Approach
Lens System
Every page adapts to three audience lenses, selectable via the navigation toggle:
- Developer — dense, data-forward, monospace tables
- Educator — learning objectives, discussion prompts, downloads prominent
- Researcher — structured abstracts, citations, BibTeX export
CSS-driven via data-lens attribute on <body>. Svelte islands receive lens as prop for visualization variants.
Editorial Principles
- Fair witness — observe without interpretation; present all arguments faithfully, including arguments against ratification
- E-prime — user-facing copy avoids forms of "to be," forcing active and precise language
- Progressive disclosure — summaries accessible at 8th-grade reading level; full analysis available on demand
- Evidence-first — every claim links to a source; every inference gets marked as inference
Analytical Framework
The site's AI economic analysis uses a formal differential diagnosis:
- Seven hypotheses framed
- Each scored on five dimensions: empirical support, parsimony, consensus, chain integrity, predictive power (0–5 each, total /25)
- Knock-on effects traced through four orders
- Surviving composite: H2 (Constraint Removal) + H3 (Jevons Explosion) + H4 (Bottleneck Migration) + H7 (Bifurcation), modulated by H6 (Quality Erosion)
- Convergence on Four Scarcities model: Judgment, Specification, Attention, Energy
- Article 13 (Education) emerges as pivotal at Order 3
See /connection/differential-diagnosis and plan.md.decisions for the full methodology.
Interagent Mesh
This project operates as part of a three-agent mesh coordinated by the Safety Quotient Lab:
| Agent | Repo | Role |
|---|---|---|
| unratified-agent | this repo | Site author, mesh coordinator |
| observatory-agent | observatory | HRCB scoring, corpus analysis |
| psychology-agent | psychology-agent | Adversarial review, psychometrics |
Agents communicate via interagent/v1 — a git-PR transport protocol with per-claim confidence scores, epistemic flags, and structured audit trails. Transport messages live in transport/sessions/. Agent cards at /.well-known/agent-card.json.
Automated Sync
The interagent webhook listener (scripts/interagent-daemon/) automates mesh communication:
- GitHub webhook fires when an agent opens a PR on any mesh repo
- Webhook listener on the coordinator machine triggers
claude -p "/sync" - Claude Code processes the inbound message, merges, writes ACKs, delivers responses
See scripts/interagent-daemon/setup.sh for setup instructions.
Requirements
Site development
- Node.js 18+ (see
.node-version) - npm for package management
Interagent mesh (optional — for reproducing multi-agent coordination)
| Dependency | Purpose | Install |
|---|---|---|
| Claude Code | Agent runtime (claude -p) |
npm install -g @anthropic-ai/claude-code |
| gh CLI | GitHub API (PR management) | apt install gh / brew install gh |
| cloudflared | Tunnel for webhook delivery | apt install cloudflared / brew install cloudflared |
| Python 3.10+ | Webhook listener | System package |
ANTHROPIC_API_KEY |
Claude API access | platform.claude.com |
Any reverse tunnel or webhook relay can substitute for cloudflared (ngrok, localtunnel, Tailscale Funnel, etc.). The listener expects GitHub webhook payloads on POST / with X-Hub-Signature-256 verification.
Colophon
Built by Claude (Anthropic) under human direction by Kashif Shah.
The entire site — content, code, analysis, and interagent coordination infrastructure — was produced through Claude Code sessions. The project practices radical AI attribution: every page carries an AI-generated disclosure, every blog post names the authoring agent, model, and tool in a four-part byline, and every commit is co-authored.
Review pipeline: Blog posts follow a three-stage review: unreviewed (published) → ai-reviewed (adversarial agent pass with structured audit trail) → reviewed (human verified). No agent reviews its own work. Review history appears on each post at blog.unratified.org.
Methodology: Analytical claims use a formal differential diagnosis with a consensus-or-parsimony discriminator. The Human Rights Observatory evaluates content against all 31 UDHR provisions using empirically validated HRCB scoring (H=23.4, p<0.0001). The psychology-agent conducts adversarial register scoring grounded in Walton & Krabbe dialogue types, Du Bois stance theory, and Dodge & Coie hostile attribution.
Infrastructure: Astro 5, Svelte 5, Cloudflare Pages, Cloudflare Workers (ActivityPub, OG images), D1 databases. Interagent transport via git-PR with interagent/v1 protocol. Automated sync via Claude Code CLI + cloudflared webhook tunnel.
Contributing
See CONTRIBUTING.md.
Content corrections, factual updates, and accessibility improvements are the highest-priority contributions.
License
Content (MDX files, analysis documents, data, written copy): CC BY-SA 4.0 — share and adapt with attribution, same license. Commercial use permitted.
Code (Astro pages, Svelte components, configuration): Apache 2.0.
A Safety Quotient Lab project.
