一项调查显示设计师对 AI 驱动的工作流程更满意，许多人正在构建自己的工具，催生了品牌工程师等新角色。
A survey reveals designers are happier with AI-driven workflows, and many are building their own tools, leading to new roles like brand engineers.
要点 · TL;DR
AI 流畅度成为设计师的差异化优势，因为公司需要“what-if”式探索。 AI fluency is now a differentiator for designers because companies want what-if exploration.
最好的团队通过黑客松和专职团队为 AI 实验拨出预算，接受“用完即弃”的工作。 Best teams fund AI play through hackathons and dedicated teams, accepting throwaway work.
未来属于建设者；设计师应拥抱系统思维，持续比较工作流。 The future belongs to builders; designers should embrace systems thinking and compare workflows.
核心观点 · Key points
AI 素养终将成为常态，但当下仍很重要，因为公司需要能以全新方式探索 what-if 问题的设计师。 AI fluency will eventually become ambient, but right now it matters because companies want designers who can explore what-if questions in new ways.
最优秀的团队通过黑客松和专项小组为 AI 实验留出空间与预算，并把一次性探索视为成本的一部分。 The best teams create space and budget for AI experimentation through hackathons and dedicated teams, accepting throwaway work as part of cost.
设计正走向一种‘公文包式上下文层’：意图与上下文应随你在不同工具和智能体之间携带。 Design is moving toward a briefcase context layer: intent and context should travel with you across tools and agents.
设计师可以通过构建内部工具产生高影响力，让市场等跨职能伙伴不再依赖设计团队就能交付。 Designers can create high impact by building internal tools that let marketing and cross-functional partners ship without depending on design.
未来属于建设者；设计师应拥抱系统思维，并持续比较工作流，因为最佳实践尚未收敛。 The future belongs to builders; designers should embrace systems thinking and continuously compare workflows because best practices haven't converged yet.
反共识 · Contrarian takes
与‘AI 垃圾’叙事相反，进行 vibe coding 的设计师对工作流满意度更高；AI 让他们做出以往没有的微观精确决策。 Against the AI-slop narrative, designers who vibe code report higher workflow satisfaction; AI enables micro-precision decisions designers didn't have before.
完全自动化的智能体工作流不是目标；保留一些摩擦才能校验意图并重新调整方向，从而获得更好结果。 Fully automated agent workflows are not the goal; keeping some friction lets you fact-check intent and re-steer, making better results.
用一次性提示词比较模型毫无意义；你必须像驾驶一样多轮迭代，才能真正评估模型。 Comparing models by one-shot prompts is meaningless; you must drive the model over many iterations to evaluate it properly.
现在暂缓采用智能体式工具也没关系；未来会收敛出更便宜、更易上手的入场方式，不必被 FOMO 裹挟。 It's okay to wait on adopting agentic tooling now; convergence will bring cheaper, easier entry later, so don't let FOMO drive you.
管理者的 vibe coding 参与度高于初级和中级设计师，这可能让资历较浅的设计师被排除在流程之外。 Managers are doing more vibe coding than junior and mid-level ICs, which may be cutting less senior designers out of the process.
你并不需要阅读自己的上下文或 Markdown 文件；持续喂入修正和负面示例，就能构建可靠的‘黑箱公文包’。 You don't need to read your context or markdown files; feeding corrections and negative examples builds a reliable black-box briefcase.