How AI is changing the Engineering side of Type Design
Recently, I’ve been working on improving the technical quality of my fonts, 0xProto and Zx Gamut. I’m happy to share that both have now passed the fontbakery check-universal tests.
Font Bakery is a standard tool used to check if a font file is technically sound. By passing the universal profile, these fonts are now better optimized for a wide range of environments and applications.
In this update, I focused on:
- Vertical Metrics: Tightening up line-height consistency across different platforms.
- PANOSE Classification: (Specifically for 0xProto) Making the metadata more precise to ensure the OS handles the font correctly.
The interesting part of this process was how I handled the fixes. I used Claude Code (Opus 4.5) to automate almost all the technical adjustments. Instead of manually editing tables, I let the AI agent handle the "fix-and-test" loop. As long as there's a clear linter like Font Bakery to provide feedback, AI is incredibly efficient at resolving these kinds of engineering errors.
This experience got me thinking about the future of font development.
I still believe that AI has its limits when it comes to the "Art" of type design, like drawing the perfect Bezier curve or making subtle optical adjustments. That part requires a human touch and a deep understanding of aesthetics that I don't think AI will replicate anytime soon.
However, the "Engineering" side is a different story. Programming for font development mostly consists of one-off scripts for the build process. It rarely requires the high-level software architecture or complex design patterns typically seen in GUI or server application development.
Admittedly, because training data for specific tools like Glyphs.app or RoboFont is still relatively scarce, creating extensions with Claude Code out-of-the-box can be challenging. However, I’ve found that simply including the API documentation in the context window almost entirely solves this issue.
By 2026, the technical side of font building has largely become a "solved problem" through automation. By offloading these repetitive engineering tasks to AI, I’ve been able to focus more on what really matters: the design of the characters themselves. It feels like a natural evolution in our workflow, and I'm excited to see where it leads.