The Evolution of Notion’s AI Coding Assistant
In a dynamic tech environment, Notion is pioneering the use of AI in its coding processes, allowing engineers to focus on innovation rather than repetitive tasks. One of Notion’s co-founders, Simon Last, exemplifies this approach as he embraces a role as a “super IC,” leveraging AI tools to streamline development.
Since introducing its AI assistant in 2022, Notion has evolved this technology into an autonomous coding agent. This shift allows the AI to operate behind the scenes, taking over coding tasks so that engineers can concentrate on more strategic objectives. Behind this transformation lies a complex process where human engineers must write substantial amounts of code to enable the AI to function effectively.
Harnessing AI for Enhanced Productivity
During a recent code sprint at Notion, I observed the engineers engaging with a tool called Cursor, which connects them with various AI models. Most of the team favored Claude, an AI known for its prowess in code generation. Engineers would typically request the AI to draft new features or modify existing ones, quickly collaborating to debug and refine the outputs before final deployment.
The relationship between AI and engineers at Notion is remarkably symbiotic. The use of generative AI, while resource-intensive, promises significant time savings. Ivan Zhao, the CEO, embodies this ideal; he seeks to optimize time spent on tasks, but often finds himself caught in a work cycle rather than enjoying leisurely pursuits, such as his passion for jazz. This cycle raises questions about the true efficiency gains promised by automation.
My time at Notion allowed me to witness the practical applications of AI in addressing real coding challenges. For instance, I was tasked with enhancing the display of mermaid diagrams within the app. Engineers Quinn and Modi explained that while these diagrams are made of SVG files, they often compromise on clarity—unlike traditional images such as JPEGs. This oversight can lead to significant user experience issues.
Quinn shared his coding strategy: asking the AI insightful questions about existing code. This method prompts the AI to conduct its own research, which can inform better prompts from the engineers. As we interacted with Cursor, it became clear that Notion’s coding environment is highly reliant on this advanced technology. The AI would relay information about the limitations of the mermaid diagrams, including their static nature and the absence of interactive features. This insight then guides further coding efforts.
As Notion continues to refine its AI capabilities, the implications for software development are profound. The blend of human creativity and machine efficiency not only accelerates project timelines but also opens new avenues for innovation. This paradigm shift is indicative of a broader trend where AI is becoming a fundamental player in tech development, enabling teams to tackle challenges with unprecedented agility.