At 50 Years Old, Is SQL Becoming a Niche Skill?

Original Article ↗ Hacker News Discussion ↗

Is SQL Becoming Niche?

  • Many argue “no,” invoking Betteridge’s law: SQL remains foundational and widely used across back-end, data engineering, and data science.
  • A minority view: “good/serious” or “advanced” SQL is becoming niche, especially in new systems where databases are treated as dumb storage and complexity moves into application code.
  • Another framing: SQL-for-development (complex modeling, optimization) is getting more specialized, while SQL-for-access (ad hoc reads, reporting) is increasingly common and democratized.

DBAs, Performance, and Schema Design

  • Teams often treat SQL as a background skill until performance collapses; then they “suddenly want a DBA.”
  • Poor initial schema/index design leads to painful refactors when load grows.
  • There are few DBAs relative to the number of projects; those with intermediate–strong skills report very high demand and pay.
  • Typical fixes: adding/removing/consolidating indexes, restructuring “hero queries,” breaking work into smaller units, and working around ORM-generated SQL.

ORMs, NoSQL, and Abstraction Layers

  • ORMs are seen as productive for 80–90% of CRUD, but often get in the way for complex queries; many developers eventually want to “just write SQL.”
  • A lot of engineering effort goes into avoiding SQL via ORMs, JSON/REST layers, or “hipster” web databases, while a relational engine still sits underneath.
  • Some teams start on MongoDB/DynamoDB for cost/simplicity, then migrate to PostgreSQL when requirements grow; others regret NoSQL as a default and are actively moving back to Postgres.
  • Several note SQL’s expressiveness and concision are hard to beat; alternative syntaxes (ORM chains, pipelines, PRQL, QUEL) are usually less compact or familiar.

SQL in Data Science and Analytics

  • In data engineering and many analytics-heavy domains, SQL is “bread and butter,” used for hours daily and preferred over Python for data processing until genuinely outgrown.
  • Some data scientists underuse SQL, doing only extraction before moving to Excel/Pandas, which others see as a productivity and collaboration loss.

LLMs, Tools, and Skills

  • One view: LLMs already write complex SQL better than most humans, making advanced SQL less worth memorizing.
  • Counterviews report LLMs failing on anything beyond simple joins/aggregates; prompts for real-world complexity are hard to express.
  • Advice recurs: modest SQL study (books, tutorials, RDBMS docs) rapidly differentiates engineers and pays off significantly.