I'm building a restaurant ingredient database using Claude as my co-developer

I have a mild tree nut allergy, and I got frustrated by the guessing game every time I eat at a restaurant. So I started building Dishclose — a site where you can look up any dish at a restaurant and see its full ingredient list, with allergens flagged and dietary tags shown. The idea began from the point of view of food allergies, but it possibly has wider appeal for anyone interested in ingredient break-downs.

The interesting part (for this community): I built almost the entire thing by having conversations with an AI. I use Claude Code (Anthropic's CLI tool) as my primary development partner. I describe what I want in plain English, and it writes the code, runs the pipeline scripts, deploys to production, and even manages the data quality. I'm a developer by background, but I've written very little code myself on this project.

What it does

• Search any dish at ~200 restaurants across Toronto, Calgary, and Banff to see every ingredient
• Allergens flagged automatically, dietary tags (vegan, gluten-free, etc.) inferred from ingredients
• "Find a restaurant" tool for groups — enter everyone's dietary restrictions and see where you can all eat
• Community voting on ingredient accuracy, confidence scores on every data point
• Blog with SEO-targeted content about dining with allergies

The stack

• Next.js 16 (App Router), TailwindCSS v4, Prisma + PostgreSQL (Neon), Auth.js
• AI pipeline: Claude Sonnet extracts menus from restaurant websites, Claude Haiku validates the results
• Feedback loops: post-scrape AI validation, search analytics (zero-result tracking), dish flagging, restaurant request queue

What I'm struggling with

• Cold start problem — I have an MVP and some data but almost no users yet
• Getting restaurant staff to verify their data (I haven't even tried any outreach yet)
• Knowing when the data quality is "good enough" to promote confidently
• Figuring out if this is a real thing that I should try to properly manage and scale, or just a fun side project

What I'd love feedback on

• Does the product make sense? Would you use it?
• Any advice on getting the first 100 users for a niche local data product?
• Anyone else building with AI as a primary development tool — how's it going? (am I likely to hit a wall that will make this approach unsustainable?)

Happy to answer questions about the AI-assisted development workflow. It's genuinely changed how I think about what a solo developer can build.