Write your UI, not AI‑only

A developer’s blunt advice in a widely shared post was simple: write the user interface code yourself, because AI still misses too many layout details to be trusted on its own. (x.com) The argument landed in a React and Tailwind CSS workflow, where developers assemble screens from reusable components and utility classes instead of drawing fixed pages. React’s own docs describe user interfaces as components, and Tailwind says its classes are meant to be composed directly in markup. (react.dev) (tailwindcss.com) That matters because modern UI code has to survive more than one screenshot. Tailwind’s responsive system applies different classes at different breakpoints, so a layout that looks correct on one screen can break on another if spacing, wrapping, overflow, or alignment rules are even slightly off. (tailwindcss.com) AI tools are explicitly selling speed on that exact task. Vercel’s v0 says it can generate “production-ready React code” from prompts or images, and multiple newer tools now promise instant React and Tailwind components from plain English. (v0ai.dev) (reactuibuilder.ai) The pushback is arriving as AI coding use keeps rising. Stack Overflow’s 2024 Developer Survey said 62% of respondents were already using AI tools in development, and 76% were using or planning to use them. (survey.stackoverflow.co) The complaint is not that AI cannot write interface code at all. It is that front-end work includes dozens of small visual states — empty lists, long labels, mobile widths, hover states, keyboard focus, error messages, and odd data — and those edge cases often do not appear in a one-shot generated mockup. (react.dev) (tailwindcss.com) React’s own teaching model starts by breaking a design into pieces and listing each component’s visual states before wiring data through it. That is close to the craftsmanship argument in the post: the hard part is not producing markup quickly, but deciding how every piece behaves when real data and real users hit it. (react.dev 1) (react.dev 2) There is still a competing view inside the same ecosystem. Generative UI products and open-source projects such as v0 and Tambo pitch AI as a faster starting point for scaffolding screens, with developers refining the result instead of typing every class and component from scratch. (v0ai.dev) (github.com) So the split is less “AI or no AI” than where the human does the final thinking. The post’s answer was that if consistency is the goal, the last mile of React, CSS, and Tailwind still belongs to the developer at the keyboard. (x.com)