Claire Vo · 3 min read · Gen AI Tools
Spec-driven development: The AI engineering workflow at Notion | Ryan Nystrom
Read the original on lennysnewsletter.com →
Notion engineer Ryan Nystrom explains how Notion is embedding AI into its engineering workflow with custom agents, the Boxy system, and a spec-first development process that turns ideas into code via Codex and Whisper. The discussion covers patterns like auto-generated standup prereads, configuring subagents, and speeding up CI through Project Afterburner, all while highlighting the value of engineers continuing to write code. The goal is to accelerate AI-driven product delivery at scale while keeping teams productive and aligned.
Key Takeaways
- Spec-first development: dictate an idea into a spec via Whisper, have Codex format it, commit it to the repository, and let autonomous agents implement and verify it.
- Notion AI custom agents and Boxy: internal tooling to surface context and generate pull requests with visuals from Notion comments in minutes.
- Fast CI is critical for AI coding agents: reducing CI time enables rapid iteration and reliability in AI-assisted development.
- Promoting reasoning: instruct AI coding agents to defend their reasoning under pushback to improve trust and accuracy.
- Engineering leadership remains hands-on: engineering managers are still writing code, reflecting a blended leadership approach in AI-enabled teams.
Excerpt
Ryan Nystrom is a software engineer at Notion. He joined in December 2024 after Notion acquired Campsite, the team communication platform he co-founded with Brian Lovin. At Notion, he’s been a core builder of Notion AI and the Custom Agents feature launched in February 2026. He manages a team of six to seven engineers while still writing code himself, currently running Project Afterburner, a push to cut Notion’s CI time to a quarter of its current duration. Listen or watch on YouTube , Spotify , or Apple Podcasts What you’ll learn: How to build a Notion AI custom agent that…