问 HN：关于文本生成 CAD 的诚实提问。这靠谱吗？
1 分•作者: giasiara•3 个月前
我是一个程序员，不是机械工程师。我一直在关注各种文本转 CAD 项目。出于好奇，我自己也开始构建一个。深入研究了现有的项目。<p>我有一些疑问。<p>那些演示都是精心挑选的，而且选得很差。着陆页、X、领英，到处都是。<p>很多这类工具只是生成网格，然后就称之为 CAD。从技术上来说没错，但真正的价值在于 BREP（边界表示）。有些工具只是使用 openscad、cadquery、build123d 等的编码代理。<p>我做了真正的用户研究。和每天使用 FreeCAD/Fusion360/SolidWorks 的人聊了聊，他们的技能水平各不相同（初学者、专家等）。我真的付钱让他们试用文本转 CAD。几乎所有人都说了同样的话：他们用 SolidWorks 等工具的速度比任何文本转 CAD 工具都快。<p>目前的 LLM（大型语言模型）在处理多个实体或真正的机械零件时就会崩溃。一些开发者会说他们完成了复杂的东西。当然。但用图形用户界面会更快吗？<p>LLM 不擅长视觉推理。我不在乎最新的基准测试结果如何。空间和几何推理目前是一个弱点。<p>那么，文本转 CAD 现在只是胡说八道吗？真的吗？我们是不是在追逐错误的工作流程？<p>我真的想听听你们的看法，至少能让我自己找到答案。
查看原文
I'm a coder, not a mechanical engineer. I keep seeing text-to-CAD projects everywhere. Got curious and started building one myself. Went deep on what's out there.<p>I have doubts.<p>The demos are cherry-picked. Badly. Landing pages, X, Linkedin, all of it.<p>A lot of these tools just generate meshes and call it CAD. Technically true, but the real value is in BREP. Some are just coding agents that use openscad, cadquery, build123d, etc.<p>I did actual user research. Talked to people who use FreeCAD/Fusion360/SolidWorks daily, various skill levels (beginners, experts, etc.). Actually paid them actual money to try Text-To-CAD. Almost everyone said the same thing: they're faster in SolidWorks etc. than any Text-to-CAD tool.<p>Current LLMs SOTA breaks the moment you have multiple bodies or real mechanical parts. Some devs will say they pulled off complex stuff. Sure. But would it have been faster in the GUI?<p>LLMs are bad at visual reasoning. I don't care what the latest benchmark says. Spatial and geometric reasoning is a weak spot currently.<p>So is Text-to-CAD just BS right now? Really? Are we chasing the wrong workflow?<p>Genuinely want to hear your takes to at least to answer myself.