AL — Public Lab for Replayable AI Infrastructure

AL is not a framework. It is a public lab.

A living collection of primitives that make AI systems auditable, replayable, and settlement-ready once they touch reality.

This repo is structured as a game:

clone it
play it
break it
extend it

Start Here

👉 GAME.md — machine-readable game loop and current levels.

Current Level Status

Level 1 — Continuity Drill      stable
Level 2 — Receipt Replay        stable
Level 3 — Replay Oath           schema live
Level 4 — Skill Boundary        implemented
Level 5 — x402 Settlement       design-ready, not active

Quickstart

git clone https://github.com/jsonwisdom/AL.git
cd AL
chmod +x scripts/root_continuity_checkpoint.sh
./scripts/root_continuity_checkpoint.sh

Then verify the receipt:

LATEST=$(ls -t receipts/root-continuity/*.json | head -1)
python3 scripts/verify_root_continuity_receipt.py "$LATEST"
python3 scripts/verify_root_continuity_receipt.py --historical "$LATEST"
cat receipts/index.json | python3 -m json.tool

What You Should See

The drill should:

• check the repo continuity state
• emit a receipt JSON
• validate the receipt JSON
• update receipts/index.json

The verifier should emit either:

RECEIPT_CONFIRMED

or:

RECEIPT_REJECTED: <reason>

Honest failure is useful. Drift is a finding.

What You Will Find Inside

GAME.md                                game loop / levels
VERIFY.md                              public replay law
scripts/root_continuity_checkpoint.sh  Level 1 drill
scripts/verify_root_continuity_receipt.py Level 2 verifier
scripts/update_receipt_index.py        receipt registry updater
receipts/index.json                    lightweight receipt registry
docs/schemas/receipt-v0.2.json         receipt schema
docs/schemas/replay-oath-v0.1.json     replay oath schema
docs/constitutional/service-primitives.md plain-language primitives
.agents/skills/                        bounded repo-local agent skills
docs/forensic/                         forensic memory
Dockerfile.replay                      replay chamber

Level 4 Skill Witnesses

• witness-replay
• receipt-audit
• canonicalization-check
• anchor-readiness

Core Loop

operation
→ receipt
→ index
→ replay verification
→ oath
→ optional settlement

Current Doctrine

No witness, no claim.
No receipt, no ratification.
No replay, no settlement.

Settlement is downstream from receipts. Payment does not create legitimacy.

How to Play

1. Run Level 1.
2. Verify the receipt with Level 2.
3. Report the first honest failure or success.
4. Extend a skill, receipt type, or verifier.
5. Keep claims bounded by what replay proves.

What This Is Not

This is not a token launch. This is not a finished protocol. This is not a claim of universal truth.

It is an operational lab for making AI-adjacent systems more replayable, auditable, and honest about their own state.

Operating Principle

No irreversible gods.
Only recoverable continuity.

Clone → run → verify → extend.