Inside Owen A and Owen B.

Two different development paths

The public architecture atlas puts Owen A and Owen B beside each other. A explores replaceable specialist AI systems around a continuing identity. B develops a nature-inspired native learning loop. Each section can be opened to see what exists, its connections, the supporting evidence and the limits of the result.

Owen A: a staged specialist package

The reviewed A package contains coordination, resource regulation, speech contracts, proposal provenance and source-continuity checks. It reports 36 fixture tests and seven model-free Windows worker cases. The package has not established live-model qualification or production promotion. Its accepted A source and memory binding remain unresolved, and visual and motion specialists retain their research or planned status.

Owen B: correction and retention

The latest reviewed B report records 19 native events and 790 visual teaching pairs in the new native school, rather than the entire historical autobiography. Bounded right-of corrections, offset and distractor transfer, and retention across a fresh process extend earlier left-of work. A retention check occurred about 58 minutes after teaching. Engineered instruction handling and guards remain part of the setup; general English comprehension is unproven.

Retina-inspired processing is present, the sparse cortical learner remains shadow-only, and replay, the additional taxel adapter and advanced sensory expansion remain disabled. These distinctions are visible in the maps.

A record that can grow

The diagrams read a versioned architecture snapshot. Reviewed changes can be published as data, so new sections appear without rebuilding the website. This is a conceptual account of the evidence, not a live neural-activity display.