Do Data Fundamentals Still Matter in the Age of AI? - Freestyle Friday (April 10, 2026)
TL;DR
Joe Reis says fundamentals are still “gravity” — borrowing Bill Inmon’s phrase, he argues first principles in data and architecture don’t disappear just because AI, new tools, or delivery pressure make them inconvenient.
“Just ship it” works until it doesn’t — Joe compares vibe-based architecture to building a house on a Thai hillside without a geological assessment: it might look fine until monsoon season turns it into a mudslide.
Data teams often inherit cargo-cult patterns instead of understanding them — he calls out how people skim Ralph Kimball, hear “star schema” from peers, and apply dimensional modeling by tribal knowledge rather than by understanding when it fits.
Scale exposes whether you actually know the fundamentals — citing practitioners who built warehouses at LinkedIn and Uber, Joe notes they eventually hit the limits of Kimball-style approaches and had to understand the underlying body of knowledge well enough to move beyond it.
AI won’t kill data engineering; it will increase the need for it — Joe predicts more refactoring, more systems work, and even “agentic data engineers,” with humans moving up a layer into orchestration while still needing to know what’s happening under the hood.
The market still rewards foundational books — he points to the sustained Amazon rankings of Fundamentals of Data Engineering, Martin Kleppmann’s Designing Data-Intensive Applications, and Kimball’s work as evidence that practitioners still want durable knowledge.
The Breakdown
Broadcasting from the Thai Hillside
Joe opens from his family’s place just north of Phuket, half-joking that he’s choosing to sweat outside because the view is too good to stay indoors. The casual setting becomes part of the theme: he’s literally sitting in a place that changes slowly while talking about what in tech actually lasts.
“Fundamentals Are Gravity”
The spark for the episode was a LinkedIn reply arguing that teams don’t really use fundamentals anymore because they’re too busy picking tools and shipping. Joe pushes back with Bill Inmon’s framing: some people build careers around vendors and tech stacks, while “fundamentalists” focus on what changes much more slowly — and that stuff still exerts gravity.
The Seduction of Practice Without Theory
Joe is sympathetic to the pressure to deliver; he says he’s been thrown into plenty of situations where you just have to build before you fully understand what’s under the hood. But his point is blunt: skipping theory feels efficient in the short term, and then one day you realize you’ve been flying blind.
The Mudslide Test for Architecture
His best analogy lands hard: wanting to build a house on the hillside behind him is not the same as being able to. You’d first bring in a geologist or geological engineer, because otherwise monsoon season might turn your dream house into a disappearing act — and he says plenty of company architectures are built exactly that way, from “vibes” and hearsay.
Kimball, Tribal Knowledge, and the Limits of Scale
Joe says people often treat data modeling as inherited doctrine: they read the first chapter of Ralph Kimball’s Data Warehouse Toolkit, hear “facts and dimensions,” and run with star schemas. He contrasts that with conversations he had in San Francisco with practitioners who scaled warehouses at LinkedIn and Uber, where old patterns stopped working at massive volumes — proving that you need to know a method deeply before you can know when to outgrow it.
Why Foundational Books Keep Selling
He points to the staying power of books like Fundamentals of Data Engineering, Martin Kleppmann’s Designing Data-Intensive Applications, and Kimball’s classics, all of which still rank consistently on Amazon. To Joe, that’s not vanity; it’s market evidence that people still want first-principles knowledge, even in a field obsessed with the next tool.
AI Raises the Bar, Not Lowers It
Joe dismisses the idea that AI makes data engineering, modeling, or architecture irrelevant. His bet is the opposite: AI creates more systems to refactor and more infrastructure to build, so engineers move up one abstraction layer into orchestration and system design — but, invoking Joel Spolsky’s “law of leaky abstractions,” that only makes understanding the internals more important.
A Quick Tour Schedule and More Data Modeling Ahead
He wraps with some travel chaos and event plugs: Agentic Analytics Summit from Cube on April 29, Data Innovation Summit in Stockholm on May 6–8 with a keynote on May 7 and workshop on May 8, and Current in London around May 18–20. Both Stockholm sessions tie back to the same thread of the episode: data modeling in the age of AI, and specifically his upcoming work around “Mix Model Arts.”