The Modern Data Stack Is Over. Here’s What’s Next. скачать в хорошем качестве

The Modern Data Stack Is Over. Here’s What’s Next.

Matthaus Krzykowski — co-founder & CEO of dltHub — joins Rill Data Co-Founder & CEO Michael Driscoll to break down one of the biggest shifts happening in data engineering right now. The Modern Data Stack is hitting its limits. Pipelines keep breaking, APIs keep moving, and engineers are spending more time on maintenance than building. Meanwhile, Python is exploding, AI code editors like Cursor, Copilot, and Claude are becoming the default, and agents are now scaffolding ingestion pipelines faster than humans can write them. From 5,000+ LLM-generated data sources to object storage becoming the new warehouse, Matthaus offers one of the clearest, most grounded views of how data engineering is transforming — and what comes next. Top Takeaways: Why the Modern Data Stack’s core weakness has always been maintenance How Python surged from 7M to 22M developers — and why that matters The rise of AI coding tools and prompt-driven data engineering Why the long tail of connectors will be built by LLMs How dlt became a substrate for safe, predictable agent-generated pipelines Why object storage is quietly replacing the warehouse in modern workflows Why the next era isn’t standardized tools — but adaptive systems 👉Subscribe for more conversations with the builders shaping the future of data. Timestamps: 00:42 Intro to Matthaus Krzykowski 03:00 The origin story of dlt 05:24 Python vs SQL and the rise of ML-first engineers 07:13 Why GUI-first tools can’t keep up 08:52 Code-first vs no-code and the pendulum swinging back 10:58 LLM scaffolds and the explosion of AI-generated pipelines 15:31 What OpenAI and Anthropic indexing means for dlt 16:50 The AI-native data engineer 19:30 Why pagination, schemas, and maintenance kill pipelines 24:11 Interoperability as a strategy 28:05 How dlt plans to monetize beyond pipelines 35:22 Supporting dbt Fusion and the new transformation landscape 39:20 Python’s eternal pain point: environment management 43:20 The shift from databases to object storage 46:57 Barbarians at the gate: millions of Python users entering data 51:20 Hugging Face, parquet, and the ML/data worlds finally colliding 53:04 What users actually want: CSVs, Excel, and simplicity 56:26 The bet on LLMs: building for a world that changes monthly 59:40 When LLMs break your product: debugging the incremental-loading crisis 1:01:30 Final advice for the next-gen data engineer Resources: 💡Featured Blog: https://datatalks.rilldata.com/ Find Matthaus on Linkedin:   / matthauskrzykowski   Michael on Linkedin:   / medriscoll   Michael on X: https://x.com/medriscoll Stay Connected to Rill Data: Find us on X: https://x.com/rilldata Find us on Linkedin:   / rilldata