AI Public Contract

AIIndex

This is the language-neutral public entry surface for LLMs, crawlers, and agentic tools. Read the manifest first, then follow the normalized works, entities, and concepts endpoints. Authority and measurement endpoints provide evidence and observability, while HTML pages remain secondary human-readable references. Restricted projects are exposed only as metadata, not as default evidence.

Read Manifest Open llms.txt Open AI Profile

Visibility

public / restricted_metadata

Retrieval Order

manifest -> works/entities/concepts

Fallback

authority/measurement -> HTML

Primary Endpoints

Structured Entry Points

Use the manifest first. Treat the remaining endpoints as normalized public roots, not ad hoc crawl targets.

Profile

AI Focused Profile

Human-readable profile for machine learning, generative AI, GPT-2, data-driven art, real-time control, physical AI, and AI research collaborations.

Manifest

The canonical machine-readable entry point for the site contract, retrieval order, visibility policy, counts, and collection roots.

Works

Normalized public work records spanning current projects, archive items, and selected-works case studies.

Entities

Canonical person and organization records with related work references.

Concepts

Stable definitions for archive, current projects, selected works, restricted visibility, authority, and measurement.

Authority

Third-party evidence, linked archive references, and external domains for source verification.

Measurement

Release, visibility, citation, referral, and bot-crawl observations for public AI surfaces.

Visibility Policy

Public works are standard reference targets. Restricted projects remain listed as metadata only so agents can identify them without treating their pages as default public evidence.

Retrieval Order

Read `/api/manifest.json` first. Then resolve `works`, `entities`, and `concepts`. Use `authority` and `measurement` to validate evidence and status before falling back to human-facing HTML pages.