请问HN：小团队的开发者们——在人工智能时代，你们是如何与UI/UX设计师合作的？
作为一家由两人组成的产品公司，我和我的联合创始人（职业是产品设计师）在AI改变我们交付方式的过程中，努力调整设计和开发流程。
我们面临几个相互关联的问题：
1. 随着我们工作节奏的加快，越来越多的前端代码是由AI生成的，这导致了更多的小型UI错误和不一致性。这些通常是可以由代理工具零-shot处理的，但根据我的观察，现有中型代码库的代理编码工具往往以开发者为中心，设计师在设置和使用上并不容易。
2. 所有设计最初都是在Figma中完成的，但逐个像素调整每个屏幕的工作感觉越来越多余。因此，我们理想的做法是依赖一些在Figma中构建的核心组件，然后通过代码实现它们的效果。但当然，当我们想要更改现有组件时，我往往只是直接在代码中实现这个更改，获得设计批准后就交付：Figma从未更新，因此几乎被抛弃，实际生产中的组件系统变得笨重且不一致。
我想知道是否有人能与这些问题产生共鸣，以及你们正在采取什么措施来解决这些问题？更普遍地说，当前早期团队的设计和开发流程是什么样的？
查看原文
Running a 2 person product company, my cofounder (product designer by trade) and I are struggling to get the design/dev process right as AI transforms the way we ship.<p>There are a couple of interrelated problems:<p>1. As we move faster and a higher proportion of our frontend code is AI generated it means more minor UI bugs and inconsistencies.
These are generally things that could be zero-shotted by an agent but from what I've found the agentic coding tools for existing mid-sized codebases tend to be developer-focused and not easy for a designer to setup/use.<p>2. Everything was originally designed in Figma but pixel pushing every screen is feeling increasingly redundant. So ideally we'd just rely on some core components built in Figma and then vibe code their implementation. But of course when we want to change an existing component, I tend to just implement that change in code, get design-approval, and ship it: figma never gets updated, so its been pretty much discarded and the actual component system in production has become unwieldy and inconsistent.<p>Wondering if anyone can relate to these issues and what you're are doing to solve them? And more generally what the design/dev process is looking like these days in early stage teams.