AI lineage, dataset compliance, and model provenance

Guides and notes on public dataset lineage, open-source model provenance, privacy, terms, legal tags, and AI compliance workflows.

LLM evaluation

GPT-5.6 Luna High Is Pareto-Optimal for FastCatalog’s Complex Classification Tasks

Across five internal benchmarks, GPT-5.6 Luna High delivered similarity to our gold reference dataset, on par with more expansive models, at about 5% of the normalized cost of the most advanced model used to generate that gold dataset. That result makes GPT-5.6 Luna High the Pareto-optimal default for FastCatalog complex workloads.

AI Atlas

Announcing AI Atlas

AI Atlas maps how public datasets and AI models are connected, letting you trace each item's upstream sources. FastCatalog.ai layers compliance intelligence (terms, privacy, origin) on top of that lineage so teams can spot risks fast.