CADmium: A local-first CAD program built for the browser
A new project called CADmium aims to be a local‑first, browser‑based, open‑source parametric CAD tool, positioning itself as a more accessible alternative to options like FreeCAD and proprietary cloud platforms such as Onshape or Fusion 360. Commenters are enthusiastic about its clear technical vision, Rust‑based Truck geometry kernel, and focus on modern UX, but raise concerns about hard problems like robust fillets, constraint solving, topological naming, and long‑term sustainability given the complexity of competing with mature commercial kernels. The conversation also surfaces broader themes in CAD: the trade‑offs between browser and native apps, the chronic UX weaknesses of many open‑source tools, licensing choices, and the importance of open, durable tools for hobbyists and professionals wary of SaaS lock‑in.
Overall reception
- Many commenters find the write‑up exceptionally clear, engaging, and motivating.
- Strong excitement that someone is attempting a modern, open, local‑first parametric CAD, seen as badly needed.
- Some skepticism that a small team can catch up with decades‑old commercial kernels and tools.
Comparisons to existing CAD tools
- Onshape is widely praised for UX and performance, but criticized for:
- SaaS dependency, lack of local‑first capability.
- Proprietary, high pricing, and prior changes to free tiers.
- FreeCAD is seen as powerful but:
- Very inconsistent and confusing in UX.
- Hampered historically by the “topological naming problem” and OpenCascade fillet/chamfer fragility.
- SolveSpace is praised as the only truly usable free parametric CAD by some, but lacks key features (e.g., robust fillets).
- Ondsel is described as a commercial distribution around FreeCAD with UX improvements, new assembly and configuration features, and cloud collaboration.
Technical discussion: kernels, constraints, and performance
- Truck (the kernel used by CADmium) is seen as promising: a potential open alternative to OpenCascade/Parasolid.
- Multiple comments emphasize how difficult robust fillets/chamfers and surface offsets are; success there is viewed as a litmus test for kernel maturity.
- Constraint solving:
- Debate over 2D vs full 3D constraint solvers; several argue 3D is necessary for assemblies.
- Claims that large constraint systems are manageable with modern sparse solvers; some think the article overstates performance limits.
- Existing solvers from SolveSpace and Ondsel are mentioned as usable references.
UX, open source, and learning curve
- Repeated theme: open‑source tools often have weak, inconsistent UX due to fragmented design and reluctance to enforce a coherent vision.
- Many recount frustration learning FreeCAD via poor tutorials and fragile workflows; contrast with relatively quick success in Onshape/Fusion.
- Others note FreeCAD UX and documentation have improved, with better tutorials and active design work.
Architecture & data formats
- Using the browser is debated:
- Pro: instant cross‑platform access (desktop, mobile, makerspaces) without installs.
- Con: some want native binaries; skepticism about matching native performance and integration.
- Electron is seen as a pragmatic way to get desktop integration given immature Rust GUI stacks.
- Using JSON for everything gets pushback from people who prefer strongly typed formats like STEP + schemas.
Licensing, openness, and ecosystem
- Choice of an Elastic‑style license prompts concern; some argue AGPL would be better aligned with “open” while still deterring cloud resellers.
- Broader worries that any VC‑backed or open‑core product may drift into subscription SaaS and away from truly local‑first use.
- Desire for:
- A non‑subscription, reasonably priced or truly free CAD.
- Preservation of parametric history and constraints in exchange formats (beyond today’s mostly geometric STEP).
- Some dislike reliance on Discord for project coordination and advocate more open communication platforms.