Descriptions:
Ryan Nystrom, engineering manager at Notion, joins host Claire Vo on the How I AI podcast to show how AI has restructured his team’s operations — from daily standups to spec-driven feature development. Nystrom manages a six-to-seven-person engineering team and describes using Notion AI with MCP integrations connecting to Honeycomb (observability metrics), Slack, and GitHub to automatically compile and post morning engineering briefings before he opens his laptop.
The interview dives into Nystrom’s internal “Afterburner” CI/DevX project, where his workflow started with dictating a feature description into Whisper, feeding the transcript to Codex to generate a written spec, then pointing Codex at the spec and asking it to build — largely one-shotting the implementation. He also describes configuring a Notion AI agent by screenshot-pasting a Honeycomb query and asking the agent to set up its own instructions from the image — an approach that mostly worked with minor manual adjustment.
Nystrom’s broader observation is that these automations don’t need to save hours per day to matter: reclaiming 20 minutes of daily context-switching is enough to meaningfully protect focus time for a team lead. He and Vo also contrast natural-language-driven agents against older deterministic workflow builders, noting that natural-language systems are dramatically easier to iterate and update. Valuable for engineering managers and senior developers looking to apply AI agents to team coordination, not just individual code generation.
📺 Source: How I AI · Published May 11, 2026
🏷️ Format: Interview
