Text-to-CAD

A new open-source project called Text-to-CAD is pitching a very sci-fi dream: describe an object in plain words, let an AI coding helper build it, then view and export the 3D model on your own computer. It promises a whole mini workshop in a box — generating models, previewing them in a local viewer, exporting files for printing or manufacturing, and even keeping edits neatly saved so you can tweak things later. On paper, it sounds like “just talk, and the machine makes the part.” Naturally, the internet showed up with both applause and popcorn.

The biggest mood in the comments? Excited, but deeply suspicious. One person praised these projects in general while warning that language models still seem shaky at real design work, basically saying the dream is cool but the reality is still wobbly. Another commenter immediately went into detective mode, asking where the actual engine was hiding — only to answer their own question after a little repo spelunking, in a very classic internet “wait… haha, found it” moment. Then came the sharpest jab: a user said you can judge the tool by the fact that it apparently doesn’t understand servo motors in the demo, which is the sort of comment that lands like a reality-show confessional.

Still, not everyone came to roast. One builder said they made a similar tool with Claude and were surprised by how well it could fix complicated shapes by looking at screenshots and trying again. Another pointed out the real chaos agent here is the human, noting that most people simply don’t have the words to clearly describe everyday objects. In other words: the AI may be learning design, but the crowd thinks humans are still terrible at giving directions — and that might be the funniest plot twist of all.