[virtualritz]

You can vibe code an app like this, relying on OBJ import (no editing apart from cutting/opening constraints), in possibly half a day.

If you doubt me, take, me up on it.

Sure, I have 35 years of experiences writing computer graphics code but I am certain I would just need to provide functional description input to Claude or Codex for this.

Zero architecture or deep 3D know-how.

The only challenge/interesting part is what happens with non-planar polygons (>3 vertices). I.e. deciding if they can be unrolled (approximated with a cylindrical or conical surface enough to 'work' when cut from paper that does not stretch).

You can alleviate this problem completely by always triangulating befor calculating any unfolding solution ofc (and get zero curved surfaces in the resulting paper model thusly).

The rest is rather trivial.

I'm not saying this isn't great, I just don't understand how you could ask people to pay for it, in early 2026.

[ndrsgrrr]

I can tell that is not trivial. I like to design papercrafts with my kids, and I use Blender but always wanted something on par with Pepakura but for Linux, so I decided to use Claude to build something similar. Furthermore, I started suggesting the Java/JavaFX stack because it is easier for me than JS, but I couldn't even create an STL viewer, so I let Claude decide the tech stack. It chose web/react/etc. (no surprise), the STL was loaded and presented as expected, the unfolding process was harder, and finally I gave up. Claude couldn't figure out by itself the best algorithm; the results were always wrong, and unfolding was just the first feature I wanted. My conclusion is that this is not the kind of application that can be easily resolved with vibecoding; the approach must be different, maybe AI assisting specific building parts to someone who knows exactly how the result should be with low-level detail.

[teaearlgraycold]

No you should vibe code an app like this and prove yourself. Then see if people actually use it.

[swiftcoder]

You should really try vibe coding a nontrivial 3D app before you die on this particular hill. LLMs are still really bad at spatial reasoning and coordinate systems. Like, painfully bad.

[jackling]

They're getting better, and are starting to be used in robotics: https://scenesmith.github.io/

[supermatt]

> If you doubt me, take, me up on it.

Given your confidence and the seemingly small amount of time you think it will take, this seems like something you should be proving rather than expecting others to do so.