Claude is an Electron App because we’ve lost native (tonsky.me)

In a recent piece, Drew Breunig explores the emergence of Claude, an AI application built on Electron rather than as a native app, raising questions about the state of application development in the context of rising generative AI capabilities. He argues that although large language models (LLMs) can handle about 90% of the coding process, the shortcomings in UI consistency and user experience in native apps contribute to the reliance on Electron. Breunig indicates that the native application environment has substantially deteriorated, with poor API management and inconsistent design leading many developers to favor web applications. This discussion holds significant implications for the AI and machine learning community, as it emphasizes the evolving relationship between application performance, developer experience, and user interface. While some maintain that native apps offer better performance and deeper OS integration, Breunig argues that native platforms often fail to deliver these benefits in practice. Instead, he contends that the real barrier remains a lack of care in application design and development, regardless of whether the stack is native or web-based. The article ultimately calls for a reevaluation of what constitutes good software in an era increasingly influenced by AI technologies.