Engineering
How we use AI to optimize CMS performance
A
Antonios Mantas · May 15, 2026 · 6 λεπτά
Inside the LLM-assisted query optimizer that automatically rewrites slow database queries on customer sites — and what it taught us about caching strategies.
Two years ago we noticed a recurring pattern across our enterprise customers: the slowest 5% of database queries accounted for 60% of total page load time. Most were N+1 problems hiding inside legacy plugins that nobody had touched in years.
This year, we shipped an LLM-assisted query optimizer that listens to slow query logs and proposes safe rewrites — eager loading, indexed JOINs, smarter pagination. The proposals are reviewed by a human before being committed to the customer's repository.
The headline number: a 2.4× average improvement on cold-cache renders across the rollout cohort.