This article reruns a 2023 benchmark with the latest LLMs, comparing direct SQL generation against querying through a structured dbt Semantic Layer. It finds that while text-to-SQL accuracy has jumped, a modeled Semantic Layer still delivers near-perfect, deterministic results for covered queries, making it ideal for complex or critical use cases.

• Top 2026 LLMs (Opus 4.6, Sonnet 4.6, GPT-5.3 Codex, GPT-5.2) now hit 80%+ accuracy on raw text-to-SQL, up sharply from the GPT-4 era's ~60%
• A fully modeled dbt Semantic Layer still reaches near-100% accuracy on in-scope queries, versus ~95% for a minimal semantic layer and ~60-80% for plain text-to-SQL
• Data modeling matters more than model choice: basic modeling boosted accuracy 15-20 points, while swapping between top-tier LLMs or reasoning settings only shifted results by single digits
• Semantic layers pay off for complex, messy, enterprise-scale data, while raw LLM text-to-SQL remains a fine low-setup option for one-off or small-dataset analysis

This article discusses a live session with Alejandro Aboy, a Senior Data and AI Engineer, about the evolving role of data engineers in the context of AI. Alejandro emphasizes that strong data engineering skills, particularly data modeling and orchestration, are crucial for successful AI work and shares practical steps for integrating AI into existing data workflows.

• Data modeling accounts for roughly 80% of the skill transfer from data engineering to AI engineering, according to Aboy.
• Good orchestration/workflow design can handle most AI tasks without needing complex agents.
• Precise column/model descriptions function as prompts for AI, so metadata and documentation quality directly determine AI performance.
• AI still lacks common-sense judgment (illustrated by a debugging agent missing context), so critically reviewing outputs rather than accepting them blindly remains essential.

This article explores the evolving role of data engineers over the past 50 years, highlighting their often unnoticed contributions to data infrastructure. It discusses the challenges they face, such as managing dependencies and schema changes, while emphasizing that the core problems remain unchanged despite new tools and technologies.

• The core problems of data engineering—dependencies, integration issues, schema changes—haven't changed in 50 years despite new tools like dbt, Iceberg, and cloud warehouses.
• Data engineers are invisible when things work but instantly blamed when something breaks.
• Real-time data requests are often unjustified since almost no one can explain how a 10-minute delay would actually change a decision.
• Industry growth reflects better marketing and rebranded terminology more than genuine advances in solving data management problems.

Since the inception of SQL in 1974, there has been a recurring dream to replace data analytics developers with tools that simplify the querying process. Each decade has seen innovations that aim to democratize data access, yet the complex intellectual work of understanding business needs and making informed decisions remains essential. Advances like AI can enhance efficiency but do not eliminate the crucial human expertise required in data analytics.

• SQL was designed in 1974 by IBM's Chamberlin and Boyce specifically to let non-programmers query data in plain English, yet this self-service dream has resurfaced every decade (OLAP in the 80s, semantic layers now) without ever eliminating the need for developers.
• AI can now generate SQL and build analytical models faster than ever, but it automates only the mechanical coding, not the judgment calls about which metrics or definitions actually matter.
• The real bottleneck was never syntax or tooling but the intellectual work of understanding business context — a gap AI accelerates around but doesn't close.