Scrutica
Sovereign-AI investment programs with announced / committed / disbursed spend stages and qualitative interdependence ratings.
Cloud GPU pricing normalised to $/petaFLOP-day across on-demand, spot, and reserved tiers, using dense BF16 throughput.
Each bundle is five files: a CSV (UTF-8, no BOM, ISO 8601 dates); a JSON Schema typing the columns; methodology Markdown that names sources and known limitations; a single-line SHA-256 over the CSV; and the Croissant JSON-LD descriptor (MLCommons 1.0) so ML frameworks can load the dataset without a manual schema mapping. Zenodo depositions mint DataCite DOIs that propagate onward to OpenAlex and Semantic Scholar, and the Hugging Face mirror ships in parallel — auto-generating its own Croissant on that side. CC-BY-4.0 throughout.
Bundles are regenerated by scripts/package-datasets.ts from the canonical processed JSON under data/processed/. Versioning is calendar-based (year.release-within-year), and fields are additive across versions; each refresh mints a fresh Zenodo version DOI under the same concept DOI, so a citation made against a prior release still resolves to that release.