Fallthrough Building For The Future
Sep 15, 2025
Jon Sabados, a seasoned staff software engineer at The New York Times, shares insights on building sustainable software systems that can evolve over time. He discusses the importance of long-term planning to avoid technical debt and the delicate balance between speed, cost, and quality in development. The conversation delves into the challenges engineers face, the journey from legacy systems to modern solutions, and even a skeptical take on AI's future role. Get ready to rethink your approach to software development!
AI Snips
Chapters
Transcript
Episode notes
Helix Failure Led To Memo Rewrite
- The team outgrew Helix, a distributed monolith that relied on one datastore and a cloud provider bug.
- That forced a greenfield rewrite resulting in Memo, with decoupled services and clearer boundaries.
Separate Systems By Scaling Needs
- Split similar-looking features into separate services if they scale differently (email vs push).
- Duplicate code early if domains differ, then refactor shared libraries later.
Use AI For Scaffolding, Not Finality
- Use AI tooling for scaffolding and repetitive boilerplate but review and fix its mistakes.
- Combine AI editors with familiar IDEs to keep productivity and correctness.