The Macro: CAD Software Is Overdue for a Fight
Mechanical CAD is one of those markets where the incumbents have been comfortable for a very long time. SolidWorks (owned by Dassault Systemes) has dominated desktop CAD for mid-market engineering teams since the late 1990s. Autodesk’s Fusion 360 moved things to the cloud and added a generous free tier that pulled in hobbyists and small shops. OnShape (now owned by PTC) was the first serious browser-native parametric CAD tool and proved that professional-grade modeling could run without a local install. FreeCAD exists for the open-source crowd but lacks the polish for production use.
The total addressable market for CAD is massive. Estimates range from $11 billion to $15 billion depending on how you draw the boundaries, and it’s growing as more physical products get designed digitally. But the interface paradigm hasn’t changed much. You still click, drag, constrain, dimension, and repeat. Learning SolidWorks takes months. Becoming proficient takes years. The software is powerful precisely because it’s complex, and that complexity is both its moat and its limitation.
There’s a newer entrant worth watching: Zoo.dev (formerly KittyCAD), which is building a code-first CAD engine with an API-driven approach. They raised a Series A and are going after the programmatic CAD use case. The fact that multiple startups are now attacking CAD from different angles tells you something about how much latent frustration exists in the engineering community.
The Micro: Type What You Want, Get a Parametric Model
Adam converts text descriptions into parametric CAD designs. You describe what you need, and it generates a model with real constraints and dimensions that you can edit, not just a static mesh. The “parametric” part matters because it means the output is actually useful to engineers. A pretty 3D render that you can’t modify is a toy. A parametric model that you can adjust dimensions on and feed into manufacturing workflows is a tool.
The founding team comes from design and robotics. Zach Dive did AI research and engineering at Adept (the AI agent company) and studied design at Berkeley. Aaron Li is the CPO and previously shipped designs to millions of users at VIVO, with earlier work at BMW and NIO, plus a design degree from Berkeley. Avi, the third co-founder, comes from a robotics background. They went through YC’s Winter 2025 batch.
The 10x speed claim is the headline. If an engineering team currently spends 40 hours on a CAD model, Adam is saying they can get a solid starting point in 4 hours. I’d be cautious about taking that literally across all use cases, but even a 3x improvement in early-stage design iteration would be extremely valuable. The first 80% of a design is often the most tedious part. Getting a reasonable starting geometry that you then refine is genuinely different from starting with a blank canvas every time.
Their mission statement is to “eliminate the gap between engineering and reality.” That’s ambitious, but the product direction is focused enough to be credible. They’re not trying to replace CAD entirely. They’re trying to make the first draft instant.
The Verdict
I think the timing is right. Generative AI has proven it can produce code, images, text, and music. Parametric geometry is harder because the outputs need to be dimensionally precise and physically manufacturable, but the underlying capability is catching up fast. The team’s mix of AI research, industrial design at scale, and robotics engineering covers the relevant bases.
The biggest risk is adoption friction. Engineers are deeply attached to their CAD tools. SolidWorks keyboard shortcuts are burned into muscle memory. Switching costs aren’t just financial. They’re cognitive and emotional. Adam’s best path is probably not asking anyone to abandon their existing tool, but rather sitting upstream of it. Generate the starting geometry in Adam, export to SolidWorks or Fusion 360 for refinement. That’s a wedge that doesn’t require anyone to change their core workflow.
The competitive risk from Zoo.dev is real but different. Zoo is going after developers who want to write code that generates geometry. Adam is going after engineers who want to describe geometry in natural language. Those are different users with different mental models. Both can win.
At 30 days, I’d want to see engineering teams using Adam for real projects, not just demos. At 60 days, retention and repeat usage are the metrics that matter. At 90 days, the question is whether Adam-generated models are actually making it into production workflows or getting thrown away and rebuilt from scratch in traditional CAD. If engineers keep the geometry, Adam has a real business. If they keep starting over, it’s a novelty.
