Doughnut.com

Inspiration

I wanted to make something meaningful out of my premium domain, Doughnut.com, instead of just letting it sit idle. At first, I thought about physical doughnuts, but I knew I could go further with it. So I started exploring ways to blend AI with food creativity. With tools like ChatGPT and DALL·E, the idea clicked: what if I could create an AI that invents completely new doughnut recipes and shows you exactly what they look like? That’s how this project began, turning a cool domain into something playful, useful, and genuinely fun for both creators and casual users.

What it does

Doughnut.com is an AI powered platform that lets anyone generate unique doughnut recipes and custom images. You type in an idea, wild or simple, and the AI comes back with a full recipe: dough, glaze, toppings, step by step instructions, and even a realistic doughnut image to match. The site includes a credit based system, user accounts, and a gallery of creations. It's basically a digital doughnut lab powered by OpenAI.

How we built it

I built it using a modern JavaScript stack. The frontend is React with TypeScript, styled using Tailwind CSS and bundled with Vite for speed. On the backend, Supabase handles authentication, storage, and PostgreSQL. I wrote serverless functions for AI calls and payments using Supabase Edge Functions. GPT powers the recipe generation and DALL·E creates the matching doughnut visuals. Stripe is integrated for credit purchases, using hosted checkout pages and webhook logic to add credits after successful payments. Zustand manages state, and I used libraries like React Query, jsPDF, Lucide, and Tiptap to round out the features.

Challenges we ran into

It definitely wasn’t smooth sailing. I ran into a lot of TypeScript issues, especially with session handling and test setups, that broke builds and caused Netlify deployment failures. Getting Supabase to connect reliably also took time. Environment variable headaches and async bugs were common early on. Stripe was another big learning curve, especially around handling webhooks and making sure credits updated securely without double processing. But bit by bit, I figured it all out (mostly using AI to debug, if I’m honest).

Accomplishments that I'm proud of

I’m proud of how everything connects and just works now. The AI generation flow, from prompt to recipe to image, feels seamless. I built a working admin dashboard, credit system, and PDF export tool, all tied together. Honestly, the fact that it’s live, functional, and looks clean on both desktop and mobile is a big personal win. This was my first proper full stack build from scratch, and I learned a lot doing it solo.

What we learned

The biggest thing I’ve learned is just how powerful AI really is, not just in what it can create, but how much it can help with building the thing itself. If you ask the right questions, it gives you better results. That applies to both the doughnut recipes and also when you're stuck debugging something at 2am. Tools like GPT and Grok.com have been a godsend throughout this. Whether it was fixing weird TypeScript errors, figuring out how Stripe webhooks work, or just brainstorming features, AI saved me more times than I can count. This whole project showed me how useful AI can be beyond just the flashy stuff. It’s like having a second brain that never sleeps.

What's next for Doughnut.com

I’m still exploring where to take it. There’s loads of potential: adding recipe filters (vegan, gluten free, etc), social features like comments and shares, maybe a mobile app, and better customization tools. I’ve also thought about letting people upload their own creations to build a bigger community around it. For now, I’m excited to see how people use it, and if someone out there sees more potential in it, I’m open to conversations.

Built With

• apis
• css
• dall?e
• gpt
• javascript
• react
• supabase
• typescript
• vite