Show HN: Chili3d – A open-source, browser-based 3D CAD application

Hacker News Discussion ↗

Overall reception and usability

  • Many commenters are impressed by performance, responsiveness, and the polished, “professional” look of the UI, comparing it favorably to mainstream CAD tools.
  • Several note it runs quickly in the browser and even “almost works” on mobile, though some operations (e.g., Booleans on iOS) fail.
  • Early rough edges: some dialogs and tooltips remain untranslated, tool names are oddly phrased in English, and the UI can suddenly switch to Chinese with no obvious way back.

Missing features and roadmap

  • Multiple users look specifically for constraints, sketches, and true parametric design; these are seen as essential for serious mechanical CAD.
  • The author (in thread) confirms constraints/parametrics aren’t in the current version but are planned as parametric components.
  • Users also note the absence of PMI, annotations, and drawing views; these are acknowledged as not implemented yet.
  • STL import is supported but limited (e.g., no snapping to mesh feature points).
  • People ask about programmatic or block-based interfaces; no direct support today, though others point to adjacent OpenCascade-based scripting projects.

Web vs native and education context

  • Strong split in opinion:
    • Some dislike browser-based CAD and argue serious 3D apps need unrestricted native access (GPU, memory control, CUDA, low-level optimizations).
    • Others report it runs faster than some native tools on their machines and praise “click and draw in 30 seconds” with no installs, accounts, or payment.
  • Educators working with Chromebooks are enthusiastic: browser CAD plus easy export to 3D printing greatly broadens access.
  • Counterpoint: reliance on web apps and Chromebooks may leave students clueless about file systems and traditional “real computers.”

Geometry kernels and technical complexity

  • Chili3d uses OpenCascade compiled to WebAssembly, which prompts a broader discussion of geometry kernels.
  • Commenters note how few mature kernels exist (OpenCascade, CGAL, Solvespace’s own, BRL-CAD, Parasolid, ACIS, C3D, etc.) and how difficult BREP/NURBS and geometric corner cases are.
  • Several experienced developers emphasize that building a robust kernel is “unsustainably hard”; reusing OpenCascade or Solvespace is seen as the pragmatic route.

Ecosystem comparisons and alternatives

  • Users compare Chili3d’s UI favorably to FreeCAD and wish FreeCAD had a similar clean ribbon-style interface; they criticize FreeCAD’s workbench model, cluttered layout, and visual quirks.
  • Solvespace is praised for constraint-focused modeling but described as painful when it breaks; some wish for a larger dev team there.
  • Onshape, Fusion 360, and other commercial tools are discussed around pricing, free tiers, and features like CAM and cloud integration.
  • Other OpenCascade-based or kernel-focused projects are mentioned (Truck, CADmium, Fornjot, Beegraphy, BitbyBit, replicad), often with notes that many are still immature or stalled.

AI, scripting, and interoperability

  • One commenter wonders if such a sophisticated project required AI assistance; replies push back, stressing that good engineering doesn’t imply AI.
  • A quote about future AI–CAD collaboration claims binary, non-scriptable tools are “high risk”; others argue that precise 3D geometry is hard to express in text and that major CAD systems already expose extensive APIs.
  • Some report good experiences using LLMs to generate OpenSCAD, suggesting text-based CAD plus AI is already practical in some niches.
  • Questions about browser-based workflows with FEA/CAM lead to examples of cloud-to-cloud integration (e.g., CAD to cloud FEM) and notes that commercial ecosystems often use APIs and PLM/databases rather than shell pipelines.