BIRD Bench research reveals that well-structured data models achieve 95% SQL generation accuracy — without semantic layer tools BIRD Bench research reveals that well-structured data models achieve 95% SQL generation accuracy — without semantic layer tools MotherDuck’s analysis of the BIRD Benchmark — a rigorous SQL generation test — reveals a counterintuitive truth: “AI-ready data simply means a clean data model. Well-named tables, straightforward joins, and clear column names are enough for frontier LLMs to...

💡 How the principles of Fact-Based Modeling could strengthen the next generation of metadata-driven and AI-assisted data platforms. Over the past months, much of the discussion around data modeling has focused on tools — how legacy suites compare to Git-native environments, and how AI is reshaping the way we create and maintain models. But a comment from Marco Wobben (author and long-time advocate of Fact-Based Modeling and FCO-IM) reminded me that, before any tool or automation, there must be semantic...

Why the future of data engineering lies in uniting structure with meaning In data engineering, we’ve long mastered the structure of data — tables, schemas, transformations, lineage. What we’ve struggled with is meaning. A recent piece by Michel Sauvage — “Ontology vs Data Model” — reignited an old but essential debate: Have we confused organizing data with understanding it? Sauvage reminds us that a data model organizes facts, while an ontology explains what those facts mean. A schema can ensure referential...

Using metadata, semantics, and AI to automate documentation and accelerate model design At this stage in the series, your metadata-driven framework has evolved into a self-aware ecosystem: YAML models define conceptual, logical, and physical layers. DuckDB catalogs and lineage graphs expose structure and quality. The semantic layer adds meaning, rules, and policy context. Now, the question is: how can we make this metadata work for us? The answer: by applying AI to interpret, generate, and...

Inspired by “Semantic Model vs Ontology vs Knowledge Graph: Untangling the Latest Data Modeling Terminology” by Cassi Hunt In recent years, the language of data modeling has grown richer — and more confusing. Terms like semantic model, ontology, and knowledge graph appear across architecture diagrams, vendor slides, and LinkedIn think pieces, often used interchangeably. Yet these concepts describe distinct — and complementary — layers of meaning in the modern data ecosystem. Understanding how they relate helps bridge business semantics with machine-readable...

How to connect concepts, rules, and business meaning across your metadata mesh As we scale metadata-driven modeling across teams, we eventually hit a new challenge: We know how data flows — but do we know what it means? We can trace lineage and structure across thousands of tables, yet still struggle to answer basic business questions like: What exactly is a “Customer”? How does “Revenue” differ between domains? Which policies govern “Personal Data”? To solve this, we...

How historization-first, model-driven data engineering extends classic warehousing and modern semantics. 💡 This article continues from Palantir’s Ontology, Kimball’s Star Schema, and Model-Driven Data Engineering: A Comparative View, responding to reader questions about how these approaches relate to Inmon’s 3NF and Data Vault — and how they look in practice. When I compared Palantir’s Ontology and Kimball’s Star Schema, several readers asked insightful follow-ups: “An Inmon-style data warehouse is in 3NF, with entities representing business objects and...

From proprietary semantics to open metadata — understanding Ontology as a paradigm, not a product. When I published Palantir’s Ontology, Kimball’s Star Schema, and Model-Driven Data Engineering, several readers reached out with smart questions: “Is Palantir’s Ontology a product or a modeling paradigm?” “Can you use the Ontology concept outside of Foundry?” “How does your model-driven approach compare to the Ontology itself, not just to Palantir?” Let’s unpack these — because understanding what Ontology really means in Palantir’s...

How data models evolved from descriptive blueprints to enforceable agreements between producers and consumers For decades, data modeling meant describing reality — identifying entities, defining relationships, and creating diagrams that visualized structure. But in today’s metadata-driven world, the model’s purpose has changed. A modern model is no longer just a representation of data — it’s a contract that encodes expectations, semantics, and quality boundaries. It doesn’t only describe what data looks like; it defines what must be true for data to...