The scaling-laws ledger audits 302 unique papers across 8 deep-loop sweeps (2024–2026) on frontier scaling-law claims (Hoffmann Chinchilla, Kaplan, DeepMind, Anthropic, OpenAI scaling reports, Llama 3.1, DeepSeek V3, Mamba/Mamba2, Hyena, Griffin, RWKV). The harness consists of 13 bills covering data-mixture conditioning, tokenizer-drift, cross-architecture replication, inverse-scaling, emergence-as-mirage, hyperparameter transfer, and architecture-class portability; six meta-costs; and three escape gates.
Three bills are predicted ★ empty and the empty-space hypothesis HOLDS across the 302-paper hand-arbitrated batch: Bill 5 ★ (causally-faithful scaling-law mechanism — 19 candidates, 0 clean triggers; three failure modes observed: smooth-metric-only, ≤7B-only intervention regime, no cross-mixture intervention), Bill 8 ★ (cross-data-mixture generalization — 12 candidates, 0 clean triggers; DCLM 0.13-spread across 53 mixtures; Yang-Bommasani 2025 cross-mixture audit predicted 0/N vendor claims survive cleanly), Bill 11 ★ (universal scaling-law cross-architecture — 59 candidates, 0 clean triggers, 33 explicit rebuttals; Albert Gu Mamba2 dense-Transformer scaling exponent fails 0.06–0.11 absolute on SSM; DeepSeek V3 MoE 20:1 fails 35–60% on active-parameter accounting).
Rebuttal density 25.8% (78 papers) — within target window. Cross-ledger couplings: Mech Interp Bill 11 ★ + Reasoning Bill 6 ★ + this Bill 5 ★ form a three-way star-mechanism alignment (causally-faithful mechanism inheritance is domain-invariant). Compute Governance Bill 19 (distilled-cousin half-life 3.4 months) ↔ this ledger Bill 11 ★ + Bill 13 — distillation portability ↔ architecture portability. This is the second ledger in the v0.2 wave produced through the full deep-loop methodology (8 parallel Opus research-agent sweeps + hand-arbitration), with 13 bills (not 19) and ★ at 5/8/11 (not 7/11/14/17) — bill count and star positions emerge from the actual scaling-law literature.