Generative UI doesn't make sense for startups
The article argues that generative UI—using AI to dynamically generate user interfaces—is impractical for startups due to high costs, latency, and unpredictability compared to traditional static UI, which remains faster, cheaper, and more reliable for most applications.
Background
- The article critiques "Generative UI" — the idea of using large language models (LLMs) to dynamically generate user interfaces (like buttons, forms, or entire screens) in real-time, rather than using pre-built, static UI code.
- The author argues this approach is impractical for startups because it's too slow (latency from LLM calls), unreliable (models can produce broken or shifting layouts), and too expensive vs. traditional UI frameworks like React.
- The piece reflects a broader ongoing debate in tech about where to apply generative AI meaningfully — many startups are experimenting with AI-generated interfaces, but critics say the core value of UI is consistency, predictability, and instant feedback, which LLMs can't reliably provide.
- The author, Evan, writes the "Code Craft" newsletter, which covers software engineering trade-offs with a pragmatic, often skeptical view of trendy AI applications.