# Suzanne is actually making it possible to design physical > Source: > Published: 2026-08-11 22:59:17+00:00 # Suzanne is actually making it possible to design physical If you're trying to build a real-world prototype, the biggest pain point is usually the transition from a sketch to a technical drawing. Suzanne streamlines this AI workflow by generating geometry that respects material properties and assembly requirements. It’s essentially an LLM agent for industrial design. For anyone wanting a hands-on guide to getting started, the process generally follows this logic: 1. **Concept Definition**: You input the functional requirements of the product. Instead of just saying "a modern lamp," you define the dimensions, the intended material (like aluminum or recycled plastic), and the specific utility. 2. **Iterative Generation**: The AI proposes several structural iterations. Because it's focused on manufacturing, it flags potential failure points or areas where the wall thickness is too thin for standard injection molding. 3. **Technical Refinement**: You tweak the parameters. Since the tool integrates with manufacturing constraints, you can adjust the "manufacturability score" to see how changing a curve might lower the production cost. 4. **Export and Deployment**: Once the design is locked, you export the files in formats compatible with standard CNC or 3D printing software. The real value here is the reduction in "back-and-forth" between the designer and the engineer. Usually, a designer creates something beautiful, and the engineer tells them it's impossible to build. By baking the manufacturing rules into the prompt engineering phase, Suzanne cuts out those wasted cycles. It's a practical tutorial in how generative AI is moving beyond pixels and text into the realm of atoms. For those of us who have struggled with the steep learning curve of professional CAD software, this feels like a much more beginner-friendly entry point into hardware prototyping. It turns the design process into a conversation rather than a manual battle with anchor points and constraints. The efficiency gain is obvious when you look at the time it takes to go from a raw idea to a valid STEP file. We are moving toward a world where the barrier to creating a physical product is no longer the ability to use complex software, but the quality of the initial intent and the constraints you set for the AI. [Backflip AI can turn a 3D scan into a parametric CAD model in 2d ago](/en/news/5636/) [Next Claude Code actually solved a math problem I'd been stuck on for →](/en/news/5956/)