Ryan Nystrom, a software engineer at Notion, has been leading the development of Notion AI and its Custom Agents feature. In an effort to streamline processes, he's working on Project Afterburner, which aims to reduce Notion's CI time to a quarter of its current duration. This initiative is crucial as fast CI has become essential in the age of AI coding agents.
Notion's internal system, "Boxy," allows engineers to @mention Codex from within Notion comments and receive a full pull request with screenshots in under 20 minutes. This enables seamless collaboration and reduces meeting prep time. Nystrom emphasizes the importance of engineering managers and senior executives continuing to write code, as it keeps them connected to the development process.
The use of AI coding agents has transformed the way developers work, making meetings more efficient and reducing burnout. Notion's Custom Agents feature allows users to build custom agents that can perform specific tasks, such as auto-generating daily standup pre-reads by pulling data from Slack, GitHub, Honeycomb metrics, and yesterday's meeting transcript.
Tools mentioned include Notion AI, Notion Custom Agents, Codex (OpenAI), Claude Code (Anthropic), Honeycomb (observability + MCP), Whisper (OpenAI voice transcription), Slack, and GitHub.
