https://www.lkranz.com/blog/ Blog posts by Luke Kranz, Full Stack Data and Software Engineer. https://www.lkranz.com/blog/who-becomes-the-product-engineer https://www.lkranz.com/blog/who-becomes-the-product-engineer Thu, 05 Nov 2026 12:00:00 GMT A feature that took nine weeks used to hide four weeks of product indecision. The build takes a week now, the indecision has nowhere to go, and the product manager and software engineer are being pushed into one job. https://www.lkranz.com/blog/is-ai-actually-making-your-engineers-more-efficient https://www.lkranz.com/blog/is-ai-actually-making-your-engineers-more-efficient Tue, 06 Oct 2026 12:00:00 GMT Writing code got faster. Deciding what to build did not. The first of three posts on AI and engineering productivity: why engineers feel faster and companies cannot find the gains. https://www.lkranz.com/blog/semantics-of-semantic-layers https://www.lkranz.com/blog/semantics-of-semantic-layers Sun, 27 Sep 2026 12:00:00 GMT Five models built a semantic layer from the same warehouse, then answered 30 finance questions against each one. The layer decided the answer almost every time. https://www.lkranz.com/blog/talk-to-your-data-harness https://www.lkranz.com/blog/talk-to-your-data-harness Fri, 25 Sep 2026 12:00:00 GMT No router I tested beat running Sonnet 5 on everything, and prompt caching saved far more than routing could. https://www.lkranz.com/blog/running-coding-agents-with-herdr-and-agent-board https://www.lkranz.com/blog/running-coding-agents-with-herdr-and-agent-board Sun, 20 Sep 2026 12:00:00 GMT I run Claude Code, Codex, and Muse side by side from a local kanban board that starts agents in Herdr tabs and shows what each one is doing. https://www.lkranz.com/blog/running-jev-in-snowflake https://www.lkranz.com/blog/running-jev-in-snowflake Sat, 19 Sep 2026 12:00:00 GMT Jevflake is a dbt package that calls TypeSafe's Jev model from inside Snowflake. Examples of each question type, what it costs, and how it compared with AI_CLASSIFY and AI_FILTER on the same rows. https://www.lkranz.com/blog/move-everything-to-an-open-table-format https://www.lkranz.com/blog/move-everything-to-an-open-table-format Sat, 19 Sep 2026 12:00:00 GMT Iceberg, Delta, and DuckLake are all strong. The hard part is that every engine reading a table has to support the same features, so pick the format by what your slowest reader can handle. https://www.lkranz.com/blog/snowflake-cortex-in-practice https://www.lkranz.com/blog/snowflake-cortex-in-practice Wed, 14 May 2025 12:00:00 GMT A year after Cortex AI went GA: LLM functions in SQL earn their keep on classification and summarization, Analyst needs a semantic view to be trustworthy, and exact numbers stay in SQL. https://www.lkranz.com/blog/dbt-mesh-and-metricflow-semantic-layer https://www.lkranz.com/blog/dbt-mesh-and-metricflow-semantic-layer Wed, 20 Mar 2024 12:00:00 GMT Coalesce 2023 gave us dbt Mesh for cross-project references and a rebuilt Semantic Layer on MetricFlow. I split our monolith, and metrics defined once finally beat metrics defined everywhere. https://www.lkranz.com/blog/go-for-data-pipelines-that-have-to-run https://www.lkranz.com/blog/go-for-data-pipelines-that-have-to-run Thu, 16 Nov 2023 12:00:00 GMT Python wins for exploration and dbt owns the warehouse, but the ingestion service that pages you at 2am is written in Go: one binary, explicit errors, and goroutines for fan-out.