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

The adoption of AI agents has transformed workflows, enabling end-to-end implementation and verification of systems through tools like Cloud Code, introduced in April of last year. While agents enhance system robustness and ambition, improper use risks inefficiency ("slop"), necessitating rigorous architectural design and verification processes. Team structures have shifted, with tools amplifying individual productivity but exacerbating performance disparities between average and top engineers. Smaller teams, such as "Tiger teams," face growing skill gaps, as elite engineers leverage tools to achieve orders of magnitude higher efficiency than peers. Culturally, the environment has become more chaotic and innovative, with increased prototyping and agent-assisted code reviews, though safety measures like controlled data access remain critical. Notions agents, now including personal and custom tools, allow users to automate tasks like database creation and report generation, while future goals focus on autonomous agent capabilities, such as self-deploying integrations, and leveraging open-source models for cost-effective AGI development.

Notion positions itself as a neutral hub for AI models, competing with larger platforms by offering access to diverse models without vendor lock-in. Its API redesign prioritizes agent efficiency, using markdown-compatible formats and SQLite databases, informed by empirical testing and user research. Internally, Notion fosters agent adoption through workshops and onboarding, emphasizing trust-building as users transition from manual oversight to relying on agents for tasks like email triage and feedback routing. However, challenges persist, including occasional agent failures requiring manual intervention and the need for new tools to manage agent interactions with workflows and databases. The role of developers is evolving from direct coding to supervising agents, with humans acting as "outer verifiers" to ensure accuracy. The future envisions a shift toward humans managing AI systems, embedding these tools into collaborative platforms like Notion to enhance productivity and reduce cognitive load.