From Coder to Manager: Navigating the Shift to Agentic Engineering with Notion Co-Founder Simon Last

The latest episode reveals how Notion has evolved from a simple writing assistant into a platform that hosts autonomous AI agents capable of writing code and constructing integrations. Simon Last explains that the breakthrough hinges on robust semantic indexing and retrieval across disparate sources such as Slack and Google Drive. Overcoming this technical hurdle required a new data pipeline that can understand context and surface relevant information instantly. By turning raw documents into searchable knowledge graphs, Notion enables agents to act without constant human prompting, reshaping the core value proposition of the product.

Internally, Notion adopted a six‑month rewrite cycle that let coding agents take over large portions of the product’s development. These agents not only generate UI components but also refactor backend services, effectively allowing the company to build itself with its own AI. Simon notes that this shift has altered team dynamics: engineers now act as overseers, defining objectives and monitoring agent performance rather than writing every line of code. The result is a productivity model where humans manage a swarm of agents, accelerating delivery while reducing repetitive manual effort.

Looking ahead, Notion is positioning itself as the ‘Switzerland’ for AI models, offering open APIs that let external developers deploy custom agents on the platform. By exposing standardized endpoints, Notion enables businesses to embed model‑agnostic agents directly into their workflows, turning the workspace into a model hub. Simon shares his personal agentic workflow, where he delegates routine research and summarization tasks to bespoke agents, freeing time for strategic decision‑making. This paradigm shift signals a broader move toward agentic engineering, where managers focus on orchestrating intelligent assistants rather than micromanaging code.