正在打开每日关注
正在打开每日关注
以饮食记录 App 为例,主张 AI 产出能否出彩,取决于人的审美判断、提示质量与持续打磨。
I’m declaring war on AI slop design. If you want your app to stand out, read this - I’ll personally fix your design for free. This insanely overdesigned calorie tracker went viral on X. I built it with Claude Code in a week. I didn’t touch a line of code. No MCPs/skills. Just careful prompting. Now I use it every day, and it’s in review for the App Store. In one week! People will call it slop in the replies. It doesn’t matter - it’s still my vision, not AI’s. It still got 700k+ views. It’s fun. Users like to play with it and show it off. THAT’S what matters for your app to be successful. AI doesn't suck at design IF you use your human brain to ask it for the right things. You have to put in the work. If you do, you capture what no other app has: your taste. The good news is you can learn this fast. You don’t need experience. You barely need any tools. I’m sharing everything about how I build delightful apps, and how you can too. If you have a topic you want me to cover, reply below or DM me. Until then, I'll lend you my human brain: drop a screenshot or video below, I’ll tell you exactly what I'd prompt to improve your design (follow me so I can DM) 👇
是什么
这条推文并非单纯展示“Claude Code 一周做出 App”,而是在回应 AI slop 争议:作者认为生成式工具负责加速实现,产品能否有辨识度,取决于人是否把品味、目标和判断写进提示并持续校正。截图里的饮食记录器是这一观点的可视化样本,把记录行为做成有情绪、有叙事的日常体验。
核心设计
界面以大标题、留白、珊瑚色背景和柔和食物图像建立鲜明语气。热量进度与三项宏量营养素先给出即时反馈;“had pancakes and a latte”等聊天气泡承接自然语言输入,随后以图片、名称、描述和营养值呈现结构化结果。“main-character energy”式文案又把数据变成生活日志。第二帧按日期展示历史记录,说明视觉表达同时服务于持续浏览。
关键洞察
可复用的方法不是堆提示词,而是先定义视觉语气、交互节奏、内容反馈和验收标准,再让 AI 执行、对照和迭代。趣味与可分享性有助于首次使用和传播,但真正的产品价值仍来自低摩擦输入、可信数据与清晰状态。
边界与风险
“700k+ views”或“用户喜欢玩”不等于留存、准确性或可访问性;过度装饰也可能掩盖核心任务。由于没有代码、测试和真实用户数据,不能据此证明一周开发质量。对资深工程师而言,最值得借鉴的是把个人审美翻译成可验证的产品约束,而不是把 AI 生成本身当成设计能力。