Vertical-Driven Architectures: reference architectures for physical AI, November 2026 release

The November 2026 release of the Vertical-Driven Architectures dataset: ten complete system designs for physical AI in physical operations (energy and utilities, maritime and ports, heavy industry and construction, oil and gas; United States, Saudi Arabia, Pakistan), as five tables. designs.jsonl holds one row per design with its summary, DOI and counts; objects.jsonl holds 136 typed ontology objects with properties, status vocabularies, anchor systems and typed links (format hyper-ontology/1); models.jsonl holds every model and hardware choice with its reason; costs.jsonl holds the three-year cost lines from each Appendix A; fulltext.jsonl holds the full text of every paper. README.md is the dataset card. Every design was reasoned on Praxis, CodeNinja's platform for designing physical AI systems, and ships an object model for Hyper Ontology. Operators are described by class, never by name. The dataset grows daily on Hugging Face; this record is the citable monthly snapshot, and later releases appear as new versions of this record. Live dataset: https://huggingface.co/datasets/CodeNinjatools/vertical-driven-architectures. Papers: https://codeatoms.ai/. By CodeNinja. CC BY 4.0. Every published design is listed among the CodeNinja Atoms sovereign AI reference architectures, with its sector, country, compute and three-year cost, in one table.