How the world's leading engineering teams are tackling legacy code, eliminating technical debt, and modernizing their systems — with measurable results.
Spotify's Honk agent automated the migration of ~1,800 downstream data pipelines away from deprecated datasets — spanning three different pipeline frameworks (BigQuery Runner, dbt, and Scala-based Scio) — saving an estimated 10 engineering weeks. A concrete case study in using AI agents to do the migration work that accumulates and never gets prioritized.
Slack's legacy notification system had accumulated four conflicting preference models over years, hidden coupling between delivery methods and content filtering, and inconsistent cross-platform state. The team rebuilt from scratch — unifying the fragmented system into one coherent model — eliminating user confusion and increasing settings engagement five-fold post-launch.
Shopify used AI agents in autonomous optimization loops to tackle CI/CD toil that engineers correctly deprioritize — achieving 65% faster build times and 53% faster parsing. A practical account of applying research-style automation to the unglamorous infrastructure debt that quietly drags down every team's velocity.