Explicit Measures Podcast 492: The 2025 Fabric Made Up Awards
Jan 8, 2026
Mike and Tommy present the whimsical Fabric Made Up Awards, poking fun at quirky features. They debate categories like 'Most Likely to Be Disabled by IT' and discuss the pros and cons of data notebooks versus traditional methods. Large language models enhance coding, but caution is needed. The duo highlights the importance of governance in future workloads and critiques overwhelming aspects like UIs and complex billing. They also share insights on performance tips, innovative knowledge tools, and the challenges of GitHub in the Fabric landscape.
AI Snips
Chapters
Transcript
Episode notes
Notebooks Are The Go-To Ingestion Path
- Notebooks (PySpark or T-SQL) are the fastest and cheapest ingestion method in Fabric compared to pipelines, dataflows, and copy tasks.
- Marco Russo's tests and the hosts' experience confirm notebooks outperform other options in speed and CU efficiency.
Use LLMs, But Validate And Watch CU Costs
- Use large language models to accelerate PySpark and notebook development but always validate generated code.
- Prefer browser Copilot for drafting to avoid CU charges from Fabric's built-in Copilot when possible.
Convert CSVs To Parquet For Performance
- Converting CSV to Parquet once yields substantial performance gains when reading from a lake.
- Parquet's columnar format often beats repeatedly reading CSVs or non-optimized formats in pipelines.