I built a parametric ceramics design platform without learning a single 3D modeling tool. Four specialized generators: dinnerware, vessels, handles, slip casting molds. The whole thing runs in a browser and exports print-ready STL files. It's called Playground Ceramics. Give it a spin.
Two years ago, that was impossible. Now it's Monday.
What I was actually doing before
I was struggling with Shapr3D. It's a beautifully designed tool — genuinely good UX, as far as 3D modeling goes. I watched the tutorial videos. I practiced. I got far enough to produce rough geometry and then hit walls I couldn't climb. I hired someone off Fiverr who knew the tool to help clean up my files. But my life is busy. I work hard and when I have time I want to spend it with my kids, my wife, ceramics, or things that make me feel like I'm moving forward. Learning a 3D tool was none of this. I was frustrated.
Why ceramics, why now
I'm a ceramicist. I love throwing and have been doing it since I was 15. I also love slip casting — the technique where you pour liquid clay into plaster molds and the plaster draws out moisture to form a shell. It's a modern step introduced into an ancient practice. Check out Edith Heath to learn about the master of the process.
What drew me to slip casting is the collision between that ancient tradition and a genuinely new capability: using 3D printing to define the forms that become slip casting molds. You design a vessel digitally, generate a multi-part mold from it, 3D print the mold, and cast in porcelain. The craft stays analog. The design tool goes digital. I'm inspired by the likes of Turn Studio and Hammerly Ceramics, who have masterfully refined this process.
I know ceramics deeply. I don't know code deeply. I never have. But I knew exactly what tool I needed. I just had no way to build it.
So I suffered through learning Shapr3D, all the while dreaming of a design tool that would help me just accomplish my goal. When MCPs became a thing, I wrote into the Shapr3D forum asking them to implement it so I could speak to the app. I waited. I dreamt. And then over the winter break of 2025, everything changed.
What changed
Andrej Karpathy coined "vibe coding" in February 2025 — describe what you want, AI writes the code. Y Combinator reported that 25% of its Winter 2025 startups had codebases 95% generated by AI. Microsoft documented a business professional with no coding background building a custom expense management app in two hours through conversation.
The label doesn't interest me. What interests me is that the person closest to the problem can now build the solution.
How it actually worked
I used Cursor and Claude Code with the latest Opus model available in December. The process was iterative and conversational. I described what I wanted, reviewed what it produced, adjusted, repeated. When something visual was wrong in the 3D viewport or the UI, I'd take a screenshot in Figma, annotate exactly what needed to change, and feed that back to the AI. Marked-up images turned out to be more precise than paragraphs of description. Vibe-coding can be addictive. There is a nifty little dopamine induced feedback loop produced when iterations yield results. I was hooked.
Why I did it
Four reasons, in order of stubbornness.
One: I didn't have time to learn a new complex 3D modeling tool. I had fought the battle with Shapr3D and lost. My knowledge is in ceramics, software, and product design, not in navigating someone else's interface. I wanted to stay in my domain.
Two: I wanted to get my hands dirty with AI-assisted creation firsthand. I wanted to build something real with it, see where it breaks, and get better at the workflow and technique. AI is a tool, after all.
Three: I'm pathologically curious and get obsessed easily. I learn by doing. These traits are perfectly suited to the rapidly evolving world of AI tools, where following your obsession is the method. See my last article How I AI for more on how I work.
Four: I think this is the greatest time to be alive as a creative. The gap between "I know exactly what I want" and "I have a working tool that makes it" collapsed in the last 18 months. From dream to reality is now a reality. Creative minds dream more than most.
It was still designing software
AI handles the code. Someone still has to decide: which parameters should be grouped together? How should the 3D viewport respond to changes? What does a good save/load flow feel like? Where do warnings about wall thickness belong?
I spent as much time on workflow and usability as I did on describing geometry. My background in software design turned out to matter. Not for writing code, but for shaping the tool into something a ceramicist would actually want to use.
AI removes the barrier of tool creation. It doesn't remove the need for thoughtful tool design.
And by the way, the tool still is far from good design. It's still raw. If you know ceramics and give the tool a spin, let me know what you think and I'll improve it.
Let's be honest about the code
Is the code clean? No. Would a senior engineer look at the architecture and wince? Definitely. I know this. The codebase grew through conversation, not through planning sprints and code reviews.
But the tool works. I can design a complete dinnerware set, generate a four-part slip casting mold with shrinkage compensation, and export print-ready files. All in a browser, all driven by parameters I understand because they map to the craft I've practiced for twenty years.
Perfect code that doesn't exist is worth less than imperfect code that ships. I'd rather have a working tool with messy internals than a pristine architecture that's still a Figma mockup. And it was so fun to build!
What it opens up
Creativity used to have three gates: the idea, the mastery of tools to execute it, and the drive to push through. AI blew the middle gate off its hinges.
The idea still has to be yours. The drive still has to be yours. But the distance between "I want a parametric mold generator with shrinkage compensation" and actually having one went from months of learning to weeks of building.
I don't know code. I never will. But I know ceramics, and that turned out to be enough.
A ceramicist builds a design platform. A musician builds a composition tool. A chef builds something that actually understands mise en place. The common thread isn't technical skill. It's domain knowledge paired with the stubborn belief that the tool you need should exist.
You already know your craft better than any AI ever will. What tool have you wished existed for years? That's your starting point.
*This article was written by me, in collaboration with Claude Opus 4.6. For the voice and tone, I asked Claude to look across the top 5 substack authors and derive principles for writing excellence to add to a skills file. I then used this skills file to edit the first draft. I'm now using Obsidian, which is a great tool for viewing and editing markdown files. I made manual edits in Obsidian, then had Opus rev again. The final version includes my hand edits.*