Caching in the Age of AI-Generated SQL

Traditional query caching assumes something that is no longer always true: that a human wrote the query, understood its behavior, and could decide whether it should be cached.

When SQL was written by hand, it was easier to:

• Identify which queries were expensive
• Determine whether results were stable enough to cache
• Define cache keys, TTLs, and invalidation logic with context

That assumption starts to break down when SQL is generated by an ORM, a framework, and now increasingly by AI coding tools:

• The developer no longer wrote the query
• The tool produces SQL from a high-level prompt or abstraction
• Queries run and often hit the database on every execution

The result is that databases absorb more read load than necessary from repetitive, expensive queries that could be safely cached if identified. Caching strategies need to adapt. Instead of relying on manual identification and static rules, systems need to observe live traffic and respond automatically.

This is the gap Readyset QueryPilot is built to close. It observes live traffic, identifies high-impact queries, and caches them automatically at the SQL layer. No manual cache keys, no TTL rules, no code changes.

For a closer look at how automatic query caching works and how QueryPilot implements it, this post breaks it down.

🔍 What We're Watching in the Database Ecosystem

AWS Engineer Reports PostgreSQL Performance Halved by Linux 7.0, But a Fix May Not Be Easy

An AWS engineer reported a major regression in Linux 7.0: PostgreSQL throughput drops to ~50% on Graviton4 due to changes in kernel preemption modes. The issue is understood, but maintainers are pushing PostgreSQL to adapt rather than reverting, creating risk as Linux 7.0 approaches release and powers Ubuntu 26.04 LTS. The impact appears limited to setups without huge pages, making them a practical workaround for now.

Database Development with AI in 2026

Brent Ozar breaks down where AI is genuinely useful in database development and where it falls short. AI is accelerating how fast SQL gets written and shipped, but the complexity and edge cases in real systems still require human judgment to manage. A practical look at what "AI writes the query" actually means for teams operating at scale.

In Case You Missed It

Introducing Readyset Shallow Cache Webinar

Last week we walked through how to cache any SELECT query without touching application code, with some great audience questions along the way. If you didn't get the chance to join us live, the recording is now live.

New on our blog:

Query Hints: Let Your Application Decide What Gets Cached: A look at how to give your application explicit control over caching in Readyset by embedding caching directives directly in SQL, enabling teams to define caching behavior at the query level.

Replication Internals: Decoding the MySQL Binary Log, Part 6 - The Query Event: Marcelo continues the series with the Query Event, the event that records statement-based changes directly in the binary log. If you're building on top of MySQL replication, this is the event you'll encounter most.

Know a database aficionado who'd appreciate this? Forward it their way.