30-Second Demo
The Origin
I struggled with traditional CAD software even after getting my first 3D printer. Tinkercad was approachable but limiting, and pro tools felt like a constant UI fight. Then I found OpenSCAD. Programmatic 3D modeling matched how I think, even if the syntax took time to learn.
I started using AI to draft the code. It could usually get close, but the workflow was clunky: copy code, render locally, take screenshots, paste them back into chat for context. At that point, building a better loop felt like the obvious next step.
The Solution
Torrify brings the whole loop into one place: chat, code editor, and real-time rendering in a single app.
- Visual feedback: Torrify sends render screenshots back to the model so it can see what changed and correct mistakes.
- Error recovery: If code fails, the error output is passed straight into the next prompt to speed up fixes.
- Smarter context: The model gets targeted examples and API guidance, which improves output quality and reduces syntax errors.
Expanding the Tech Stack
I recently added support for Build123d, a Python-based library that handles complex shapes especially well. That meant building a custom context layer to help the AI navigate a larger API. It’s still in progress, but it already unlocks new geometric options.
Future Vision
I created a custom .torrify format that saves more than
geometry. It keeps the conversation history, code iterations, and generated
images together.
Long term, I’d like to build a sharing platform for these files, similar in spirit to MakerWorld. It would create a high-quality dataset linking human intent to working code, and help drive better generative tools over time.