软性材料工程比钢铁更难 ✍ amy🕐 2026-04-22📦 4.9 KB 🟢 已读 𝕏 文章列表 文章深入探讨了软性产品工程的复杂性与技术挑战。作者指出,相比于各向同性的钢材,纺织材料因其各向异性、复杂的层压结构及粘合剂匹配问题,更难进行预测和建模。目前该领域缺乏现代化的CAD工具,主要依赖经验知识。文章认为软性工程是时尚与机械工程的交叉点,且针对该领域的AI工具开发尚处于空白期,蕴含着巨大的商业和技术机会。 软性材料工程学CAD仿真制造供应链材料科学设计纺织技术AI工具 # Fabric is harder than steel **作者**: amy **日期**: 2026-04-20T22:08:24.000Z **来源**: [https://x.com/amypretzel/status/2046350294808736064](https://x.com/amypretzel/status/2046350294808736064) ---  When I tell people I worked on softgoods at Apple, most of them assume I was a designer who picked colors and textures. The reality is that softgoods is one of the most technically demanding disciplines in product engineering, and almost nobody outside the industry knows it exists. Softgoods means anything flexible: textiles, leather, foam, films, woven and knit composites, the laminated stacks inside your shoes, the fabric on your headphones, the case on your laptop sleeve. It's the part of the product that has to bend, stretch, breathe, drape, and hold its shape across years of human abuse. And it's much, much harder to engineer than rigid parts. Why steel is easy A block of steel is isotropic. Pull on it in any direction and it behaves the same way. You can hand it to a CAD program, mesh it, simulate it, and the answer you get back is reasonably close to reality. You can specify the material with a single line on a drawing: "1018 steel, cold rolled." Done. Now take a piece of woven nylon. It has a warp (long axis) and a weft (cross axis), and the two have completely different mechanical properties. Pull on the bias (45° to both) and you get something else entirely - much higher elongation, much lower stiffness. The fabric is anisotropic, and there is no single Young's modulus you can put on a drawing. Knits are worse. They deform by loop geometry change before the yarn itself starts to load, which means they have a huge low-stiffness region followed by a sharp inflection. They creep under sustained load. They behave differently after the first wash. CAD doesn't model any of this well. The drape simulators that exist (CLO3D and similar) come from the apparel world, not the engineering world, and they're optimized for visual realism rather than predictive mechanics. The lamination problem Most modern softgoods aren't a single material. They're a stack: a face fabric, an adhesive layer, a foam or film core, sometimes a backing fabric. The stack is what gives the part its hand-feel, its dimensional stability, its acoustic or thermal properties. The stack is also where everything goes wrong. If the modulus of your adhesive doesn't match the moduli of the layers it's bonding, you get shear stress concentrations at the interface. Over time and thermal cycling, the adhesive creeps, the bond degrades, and the layers delaminate. The product looks fine on day one and falls apart at month nine. Designing a lamination stack means thinking about: - Modulus matching across all layers - Coefficient of thermal expansion mismatch (a fabric and a film expand at very different rates) - Surface energy and chemical compatibility for the adhesive bond - Cure schedule and how it interacts with downstream assembly heat - Wash and abrasion durability if the part is exposed There is no software tool for this. It is tribal knowledge, learned by ruining a lot of parts. The best of both worlds What I love about softgoods is that it sits at the intersection of two disciplines that almost never talk to each other: fashion and mechanical engineering. The fashion side gives you intuition about hand-feel, drape, color, how a material reads to a human. The engineering side gives you the language to predict failure, specify tolerances, and run a real supply chain. Both are essential. Most teams have one or the other. The brands that make great soft products - Patagonia, Arc'teryx, Nike, Apple's softgoods team, the high-end automotive interior shops - have figured out how to put both kinds of brain in the same room. That's harder than it sounds. Why this matters now AI tooling for hard goods is exploding. Generative CAD, simulation, the whole stack is getting faster and smarter every quarter. Softgoods has barely been touched. The textile industry runs on PDFs, hand-drawn tech packs, and decades-old simulation tools that nobody outside apparel has heard of. There is enormous room for someone to build the equivalent of modern CAD for soft materials, and almost nobody is trying. Variant 3D is one of the few teams I've seen actually working on this; there should be ten more. If you're an engineer looking for a discipline that's technically deep, commercially significant, and basically empty of competition, look at softgoods. ## 相关链接 - [amy](https://x.com/amypretzel) - [@amypretzel](https://x.com/amypretzel) - [Upgrade to Premium](https://x.com/i/premium_sign_up) - [6:08 AM · Apr 21, 2026](https://x.com/amypretzel/status/2046350294808736064) - [665K Views](https://x.com/amypretzel/status/2046350294808736064/analytics) - [View quotes](https://x.com/amypretzel/status/2046350294808736064/quotes) --- *导出时间: 2026/4/22 15:37:50*
视 视觉锚点:小红书封面的核心逻辑 文章深入剖析了小红书封面设计的关键——视觉锚点。作者指出,单纯追求美观不如构建清晰的信息密度和视觉锚点以提升点击率。文章总结了四种高频封面结构(冲突型、数字型、截图型、情绪型),并分析了平台的算法偏好。最后,作者发布了相应的 GitHub Skill 工具,辅助用户快速生成符合逻辑的封面方案。 技术 › Skill ✍ 波妞PONYO🕐 2026-06-27 小红书运营设计视觉锚点封面设计点击率AI工具ChatGPT方法论创作技巧
压 压进我十年设计经验的 PPT Skills,迎来大波更新 作者歸藏介绍了其开源项目 guizang-ppt-skill 的重大更新。本次更新新增了基于瑞士国际主义设计的“瑞士风”视觉风格,并接入了 Codex 的 GPT-Image 2.0 功能,实现了自动生成符合调性的配图(如胶片质感人像、流程图)。此外,Skill 现已支持一键生成小红书、公众号等多平台封面,进一步打通了从大纲到 PPT 生成再到跨平台发布的全流程。 技术 › Skill ✍ 歸藏🕐 2026-05-12 PPTClaudeCodex设计AI工具瑞士国际主义开源视觉风格
如 如何用 Claude Skill 制作高质量 PPT(附完整教程) 文章介绍了作者基于 Anthropic 的 Claude Design 功能封装的一个开源 Skill,该工具能通过对话自动生成高质量 PPT、产品宣传动效及前端网页。文中提供了 GitHub 地址,并详细演示了从安装到实战(文章转 PPT、产品文案转动效)的全过程,旨在降低设计门槛,帮助普通人快速将想法转化为高质量的可传播内容。 技术 › Skill ✍ 阿西_出海🕐 2026-04-27 ClaudeSkillPPT设计教程自动化AI工具Claude Code开源工作效率
连 连接四个 Skill 实现视频生成到成片的全流程自动化 文章介绍如何通过串联 Seedance 2.0、video-use、Remotion 和 FFmpeg 四个 Skill,让 Codex 实现从视频生成、剪辑、字幕制作到格式导出的全自动化流程,大幅提升视频制作效率。 技术 › Skill ✍ sitin🕐 2026-07-29 Codex视频生成自动化RemotionFFmpeg视频剪辑AI工具
C Codex 10大中文创作者必装Skills 文章介绍了Codex平台上10款专为中文创作者设计的必装Skills,涵盖内容写作、去AI味、多平台发布、配图生成等功能。推荐优先安装Humanizer-zh、dbskill等工具,以优化文字质量并提升创作效率。 技术 › Skill ✍ 启航🕐 2026-07-29 CodexSkill内容创作AI工具效率提升去AI味小红书公众号X平台
至 至超级个体 本文探讨了AI时代的“超级个体”概念,指出超级个体并非培训而成,而是由好奇心激发的AI Builders。文章回顾了互联网从独立软件时代到分工细化的工业化、平台化演变,认为LLM和Agent工具正在重新压缩分工,赋予个人闭环能力,大组织应调整人才策略以适应这一变化。 技术 › Agent ✍ Henry Li🕐 2026-07-29 超级个体AI BuildersAI时代人才战略职业发展LLMAI工具个人成长Closed-loop职场管理
大 大学毕业一年,试过小红书、YouTube、web出海后,我终于明白了什么值得长期积累 作者Oscar毕业一年,在小红书、YouTube和Web出海项目中不断试错。经历旅游种草变现和AI视频断流后,他意识到短期流量不如长期资产重要。他提出四个筛选方向的方法,强调区分流量与资产,利用AI但不依赖AI,通过最小闭环验证想法。 职场 › 职业发展 ✍ Oscar Growth🕐 2026-07-28 副业内容创作个人成长AI工具Web出海试错复盘长期主义
2 2026最强 AI 组合之一:Codex + Figma,不会代码也能做产品! 本文介绍了 Codex 与 Figma 结合的 AI 产品开发工作流。通过 Figma MCP 连接两者,实现了从需求梳理、设计生成到代码开发及浏览器验证的闭环。文章详细讲解了如何利用 Codex 进行需求分析和代码编写,以及如何在 Figma 中进行双向协作与设计评审,大幅降低了产品原型开发的门槛。 技术 › Codex ✍ 云析🕐 2026-07-28 AI产品开发FigmaCodexMCP前端开发设计工作流
K Kimi K3宣传图解构:Kimi Archive 风格提示词 作者受 Kimi K3 宣传视频启发,深度拆解其“Kimi Archive”档案美学风格。文章提供了一套详细的 AI 绘画提示词模板,涵盖构图、物件、质感与色彩规范,旨在指导用户生成具备高级感、克制美与收藏价值的封面视觉,适用于未来创意与设计场景。 技术 › 工具与效率 ✍ Adrian Punk🕐 2026-07-27 提示词AI绘画设计Kimi视觉风格封面设计
d dashi-ppt 我们也测了,第一还是 slide-maker 文章对比了 slide-maker 和 dashi-ppt 两款 AI PPT 生成工具。在基于 Opus 5 模型的统一评测中,slide-maker 的无图版和带图版分别获得 94.5 和 91.5 分,击败了获得 83.5 分的 dashi-ppt。评测指出,slide-maker 胜在代码生成带来的高自由度和原生可编辑性,而 dashi-ppt 虽导出效果好,但受限于模板,内容结构被强行适配,导致丢分。 技术 › Skill ✍ Martin(小马)🕐 2026-07-27 PPT生成评测slide-makerdashi-pptOpusAgent对比工具评测开源AI工具
前 前阿里P8被裁后用WorkBuddy月入17万——我扒了27个真实案例,发现赚钱的人只用了5个功能 文章通过27个真实案例,分析了如何利用WorkBuddy的5个核心功能(定时自动化、多Agent并行、Skill技能包、MCP连接器、记忆系统)实现自动化赚钱。作者指出,赚钱的人将AI当作员工用,构建自动化工作流,而不赚钱的人仅将其视为聊天工具。 技术 › Agent ✍ 沈美丽子🕐 2026-07-27 WorkBuddyAgent自动化技能包案例赚钱AI工具效率多Agent定时任务
我 我是如何用 Codex 做自己的个人IP生图工作流的 作者分享了如何利用 Codex 构建个人 IP 生图工作流的经验。通过借鉴优秀项目,将思路从拼图改为拼提示词,开发了封面图和段落图两个 Skill。文章详细介绍了风格确定、人物特征锚定、内容总结及安全区适配等优化过程,强调了 AI、代码与人在工作流中的职责分工。 技术 › Codex ✍ naiyue777🕐 2026-07-26 CodexAI绘图个人IP工作流Skill自动化设计开源项目