How about a robust CAD application built in Rust that supports designing and engineering everything from fashion to landscapes to 3D printing and more?
We are constantly looking for ways to test CodeYam and put it through its paces on challenging software. Last month, we saw Brandon Thomas’ post that he had created a vibe coded, open source alternative for Adobe software apps from scratch (further discussed on Hacker News and extensively discussed on tech Threads / Reddit / Twitter).
This inspired us to set some new goals to test out CodeYam by building some new complex app ideas.
CodeYam helps with design and feature explorations as well, though, so we did not want to just clone an existing application. Still, a complex desktop application built in Rust felt like a good exercise and test.
I’ve been exploring adding a screen porch to our house. I’ve poked around using various tools to design it out, but found it always takes quite a while to learn the tool and exercise it effectively enough to produce quality design explorations. CAD software and the various other forms of design and engineering software are almost universally like this in terms of the steep learning curve to get to a useful result as a new user.
This presents a new possibility though: what if you could build a general use design and engineering tool from scratch where one could more easily collaborate with an AI agent?
The AI agent could teach you how to use the application or, by describing what you want, control the application itself to produce the desired outcome. The more mundane aspects of bringing your vision to life could be handled by the AI in a fraction of the time.
And so CadYam, built by CodeYam, was born.
The goal was not to build the app in four days. The app is nowhere near complete, but the progress made in just four days, armed with a Claude Code Max account and an average of 6 virtual machines (VMs), each running CodeYam, seemed impressive enough to show off.
Over the course of four days and 244 commits, CadYam came to life with extensive functionality, 3,770 Rust tests, and 795 scenarios (which are a form of e2e test CodeYam uses to test, capture, and demonstrate how the application works).
This particular situation was benefitted by the fact that CAD-like tools are relatively well-understood so the AI agent was able to help explore potential features. The fact that the AI agent can also easily drive the CadYam application made manual testing and verification easier as well.
Enough of the process, though, let’s show CadYam in action. Note: this demo is from the version that was created after four days of work. CadYam has since been further developed and refined and we’re hoping to share an update on our more recent progress soon.
Our mission is to empower individuals and teams to build beautifully designed, highly maintainable software with CodeYam.