Malleable software empowers users to customize digital tools locally, shifting control from centralized developers to the people who actually use the software. While AI agents significantly lower the barrier to coding, they currently lack the refined "taste" necessary to design elegant, reusable abstractions. Consequently, understanding has become the new bottleneck; as AI automates technical execution, humans must maintain a deep conceptual grasp of their projects to drive creative direction. Effective software design requires building modular, extensible primitives—similar to Notion’s block-based system—that allow for gradual, incremental learning rather than overwhelming complexity. Drawing inspiration from historical computing environments like Smalltalk, which prioritized live, modifiable systems, developers can create tools that balance convenience with the necessity of human agency, ensuring that AI serves as a partner in the creative process rather than a replacement for human thought.
Sign in to continue reading, translating and more.
Open full episode in Podwise