From Assistance to Human Capability: A Readout-Genesis Architecture for Proactive AI, Unequal Life Conditions, and Opportunity Conversion (RG-HCA v1.1 glosa-checked edition)

Status. K0 conceptual and measurement architecture with an empirical research agenda; v1.1 is the glosa-checked edition of the 6 September 2026 v1.0 (citation lineage repaired to deposited DOIs, versions and public git anchors; priority-adjacent wording replaced by a contribution boundary; tier tags added to every boxed equation; no content change). Part of the series When AI Expands Human Potential (index DOI 10.5281/zenodo.22308201). Extends the Dialogue Conversion Protocol (10.5281/zenodo.22481928) with a proactive capability layer and sits under the macro boundary of After Labour (10.5281/zenodo.22481924) and The Human Conversion Imperative (10.5281/zenodo.22481926). Central claim. When durable human development is an agreed goal, AI need not remain purely reactive: starting from a human-owned live problem it may diagnose whether an observed difficulty is a knowledge or skill gap or a structural barrier (language, tool, resource, network, credential, permission, opportunity), generate non-equivalent capability and opportunity routes, ask for human endorsement, scaffold adaptively, withdraw decisive assistance, test unaided Human Return, and connect retained capability to world feedback and real opportunity. Non-collapse laws: Observed Difficulty ≠ Skill Deficit; Proactive Suggestion ≠ Human Goal Ownership; Credential ≠ Capability; Market Legibility ≠ Human Worth. A worked Southern-Thailand three-child scenario, child and adult governance bundles, six outcome families, a five-study programme, six rival models and removal conditions are stated. No exact human-domain quotient or calibrated universal scale is claimed. An AI assistant assisted under the author's direction; no AI system is an author or contributor.