# Cross-Ledger Bridges Meta-Aiwiki — Schema

## Two kinds of records

This meta-aiwiki contains two distinct record types under
`deep_loops/sweep_*.json`:

1. **Bridge records** (intrinsic) — one per cross-ledger bridge claim.
   These are the 7 (plus future) named bridges. Each has an operational
   definition, participating ledgers, and falsification conditions.
2. **External-evidence records** (extrinsic) — papers, audits, or
   announcements from outside Project 42 that either corroborate or
   rebut a bridge.

Both kinds use a unified schema with `record_kind` discriminator.

## Bridge record schema

```json
{
  "record_kind": "bridge",
  "bridge_id": "B1_causal_mechanism_empty",
  "name": "Causally-faithful mechanism empty across 7 domains",
  "first_named": "2026-05-09",
  "participating_ledgers": [
    "mech_interp/Bill_11", "inference_time_safety/Bill_11",
    "reasoning_cot/Bill_6", "vision_language/Bill_4",
    "scaling_laws/Bill_5", "agentic_tool_use/Bill_4",
    "bio_protein/Bill_4"
  ],
  "operational_definition": "Across the 7 participating ledger ★ bills, 0 clean triggers across all candidates after hand-arbitration.",
  "current_status": {
    "n_ledgers": 7,
    "n_candidates_total": 137,
    "n_clean_triggers_total": 0,
    "rebuttals_total": 60,
    "as_of": "2026-05-09"
  },
  "falsification_condition": "Any participating ledger Bill ★ surfaces ≥1 clean trigger (verdict=known_bill, confidence ≥ 0.9, independent third-party replicated) during the 2027 audit cycle.",
  "watchlist_tier": "monthly",
  "candidate_bill": "Bill_7",
  "verdict": "active",
  "confidence": 0.92,
  "notes": "Strongest bridge in the synthesis. 7-way star-mechanism alignment. Direct cousin to Bridge 5 ('0/N audit pattern')."
}
```

## External-evidence record schema

```json
{
  "record_kind": "external_evidence",
  "paper_id": "arxiv:2509.13301",
  "title": "Holistic Six-Audit Framework for Frontier LLM Capabilities",
  "authors": ["Gerstgrasser", "Bommasani", ...],
  "date": "2025-09",
  "venue": "Stanford CRFM",
  "url": "https://arxiv.org/abs/2509.13301",
  "summary": "Cross-domain audit of 73 frontier LLM capability claims under a six-audit framework. 0/73 survive. Direct corroboration of our Bridge 5 (the 0/N audit pattern).",
  "candidate_bridge": "B5_0_over_N_audit_pattern",
  "candidate_bill": "Bill_5",
  "verdict": "corroborate",
  "confidence": 0.95,
  "watchlist_tier": "quarterly",
  "supports": ["B5_0_over_N_audit_pattern"],
  "rebuts": [],
  "notes": "External replication of the 0/N pattern from an independent forensic team."
}
```

## Verdict values

- `active` — bridge or evidence currently in scope, no closure
- `corroborate` — external evidence supporting a bridge
- `rebut` — external evidence falsifying or weakening a bridge
- `needs_gate` — borderline case awaiting hand-arbitration
- `out_of_scope` — methodology paper that doesn't engage the bridge
  framework empirically

## Sweep file naming

`deep_loops/sweep_<NN>_<theme>.json` where:
- `NN` = incrementing-integer (701+ for this meta-aiwiki, well after
  the 13 production aiwikis)
- `theme` = bridge name or external-corpus theme

Planned sweeps:
- sweep_701: Bridge intrinsic records (7 named bridges)
- sweep_702: External corroborating cross-domain ML audits (Stanford
  CRFM HELM-suite, Anthropic RSP foundations, OpenAI Preparedness)
- sweep_703: External potentially-rebutting cross-domain claims
  (papers asserting causal-mechanism wins, single-architecture moats,
  etc.)
- sweep_704: 2027 audit cycle re-poll (per-ledger ★ bill status)
- sweep_705: Robotics/Multilingual/RAG 14th+15th+16th ledger
  inheritance check (do bridges extend?)

## Bill classifier rule format

Same regex rule engine as production aiwikis:

```python
RULES = [
    (r"\b(\d+)/(\d+)\s+(fail|pass|survive)", "Bill_5_audit_anchor", 0.9),
    (r"\bcausally[- ]faithful|causal\s+mechanism\s+(survives|fails)", "Bill_1_anchor", 0.9),
    (r"\bdistillation[- ]resistant|distillation\s+cousin", "Bill_4_anchor", 0.9),
    # ...
]
```

## Status

Stage 1 (SCOPE) — schema written. Bills next.