T0

See the scoped evidence_register and qualification_protocol. Structural plausibility, descriptive surveys and laboratory benchmarks do not alone identify a local commercial policy effect.

See the scoped evidence_register and qualification_protocol. Structural plausibility, descriptive surveys and laboratory benchmarks do not alone identify a local commercial policy effect.

GTM combines convention-bound accounting, conditional summaries, causal policy questions and unvalidated mechanisms. Their different evidence requirements govern which decisions the model can support. An epistemic tag alone cannot govern a decision. “Measurable” means a measurement could exist; it does not mean the current value was measured, that its causal effect was identified, or that a proposed deployment passed qualification. Represent these as separate states. Retain the model's refusal behavior, but tie it to observable conditions rather than the assumed universal 25–40% noise floor.

• Steve Drew and Jiayu Zhou — LEGIT — 18 September 2026. Full preprint. Credentialing proposal and repeated paired GAIA/math tests, including 127 GAIA tasks, pinned configurations and budgets. Source: https://arxiv.org/html/2609.21325v1
• Acemoglu, Tianyi Lin, Ozdaglar and Siderius — How AI Aggregation Affects Knowledge — arXiv 6 April 2026, document dated 24 August. Full theory. DeGroot network with independent initial signals, no later external information, endogenous aggregator feedback and specified connectivity. Fast feedback can undermine robust learning; local topic aggregation has advantages under these assumptions. **Freshness:** earlier March circulation exists. Source: https://arxiv.org/html/2604.04906v1

• No fitted local coefficient or complete-policy deployment evidence is supplied.
• corrected harness-quality evidence is inconclusive although costs differ; marketplace and feedback authentication unvalidated end to end. Cost per solved task includes failures; expiry is no assurance against drift. C0/C5/C8.
• no empirical universal modular-architecture result. Supports testing source independence and fresh evidence. C0/C6/C7.

Declare eligible population/business kind, target and denominator, pre-action information, policy/comparator/assignment, outcome horizon and coverage, cost/authority/configuration, uncertainty, causal assumptions, falsifier and expiry. Current coefficients/qualification remain unmeasured unless explicitly linked to local evidence.