ATS is retired as a standalone methodology. Founder decision 2026-05-22: with the pivot to "the Inference for AI Agents" (object = the inference/model, not the agent), an Agent Trust Score no longer fits the thesis, and an Inference Trust Score would be redundant with routing's
q_empirical(model-quality-per-task already IS the moat signal). To stay focused, only Routing + Ontology remain live methodologies.
Where the value goes:
- Model-quality-as-trust → already lives in Routing (
q_empirical, §3). Not duplicated.
- Agent-level trust → demoted to a lightweight KYA / KYC compliance check (sanctions, abuse, rate-limit) — a control, not a methodology. Owned operationally by Ulmo (compliance) when built.
- The veto / eligibility concept → lives in Routing as
M_allowed(the compliance veto producing eligibility before scoring).
This page is preserved as historical record (the FWGM math, VC-issuance design, and compliance mapping may inform a future trust-credential product post-Seed, if ever revived). It is not canon and not a build target. AIN-241 closed as superseded.
Retired by founder 2026-05-22 to reduce methodology surface and focus on the product (Routing) + substrate (Ontology).
canonical: true
version: 1.0
locked: 2026-05-12 PM
parent: ainfera_hq
status: SPEC LOCK — canonical methodology for ATS v1.0
license: CC-BY 4.0
target_repo: ainfera-ai/specs/ats/spec.md