Demo

Proof, not promises. Browse the ledger, inspect the pipeline, verify anyone’s. Everything runs client-side — no account, no data leaves your browser.

How it works

Every push to the AIIR repo runs this pipeline — the same one you’d use.

1
Commit lands Developer pushes to main
2
AIIR generates receipt CI runs aiir --pretty
3
Hash is verified SHA-256 of core fields → content_hash
4
Anyone can check Browser, CLI, or online verifier

Community receipt explorer

Full auditability: every receipt the AIIR project has ever generated, published through the public receipts flow. The site reads a generated community index, so feature-branch work in AIIR stays separate from the public receipt archive. When the receipts publisher updates that index, new receipts appear here automatically. The community can use this to evaluate AIIR and to help make the heuristic better.

Verification lives on the verifier page

The dedicated /verify page explains the trust tiers directly: browser verification proves Tier 1 content integrity, while signed receipts add stronger provenance. The demo stays focused on the public ledger and dogfood pipeline.

0
Total receipts
0%
Classified human
0%
Bot
0%
AI-assisted
0
Unique signals
Known detection gaps

false negatives ~195 commits classified “human” were actually authored via Copilot Chat (Claude Opus 4.6 / ChatGPT 5.4). The heuristic can’t detect editor-based AI that adds no Co-authored-by trailers.

false positives Several “AI-assisted” detections triggered because commits mention tool names (Claude Code, Duo Chat, Lovable) as features being built, not because those tools wrote the code.

Commit Subject Author Class Signals Date
Page 1

Try it on your repo

pip install aiir && cd your-repo && aiir --pretty
Read the docs → GitHub →