ChowPilot - Your AI-powered co-pilot for the kitchen

Inspiration

I love cooking — but not the admin that comes with it. I was tired of endlessly searching for recipes, building shopping lists, and repeating the same meals week after week. ChowPilot was born from a personal need: I wanted something that would make home cooking exciting again by cutting out the boring bits — and leaving the fun of creating and eating great food.

What it does

ChowPilot is your AI-powered co-pilot for the kitchen. It helps you:

Generate personalised meal plans tailored to your taste, dietary needs, and number of servings.
Choose from a range of interesting cuisines. Cook fast or go fancy with restaurant grade options.
Create smart, consolidated shopping lists with direct links to ingredients on your favourite supermarket websites (currently Australia only)
Get step-by-step audio cooking instructions, so you can cook hands-free
Learns what you like over time and suggest even better dishes every time you use it

You can try it for free, without even creating an account.

How I built it

I challenged myself to build this almost entirely using speech prompts — no traditional code, and almost no keyboard work. The stack includes:

Bolt.new – to build the web app, landing page with a slick in-browser IDE and easy integrations with other tools below
Supabase – user auth and meal plan storage and edge functions
Netlify - hosting, build pipeline and forms
OpenAI – for meal plan generation
Pica – orchestrating API connectivity across numerous services (and switching out other recipe models quickly)
ElevenLabs – for realistic text-to-speech cooking guidance
Stripe – to handle subscription billing
Entri – exploring, buying and connecting domain names
Aqua Voice - speech to text tool for prompt input

Challenges I ran into

Adapting to my own changing plans - Being able to iterate so quickly meant that I was outgrowing what I'd built just hours before.
User onboarding - Specifically allowing an anonymous user to create a meal plan and then import that into the database when they sign up for an account. In the end I used the local storage API and a series of migration steps that trigger with auth.
Streaming and caching the audio guide steps
The non-deterministic nature of AI - Getting results from OpenAI in a consistent format took more prompt engineering than expected.
Securing a fun, memorable brand with a .com domain - Naming things is hard!
Token optimisation - I really went wild in the first few sessions before getting a better feel for how to tune prompts to get the best out of the system.

Accomplishments that I'm proud of

Fully functional AI-driven app with meal planning, grocery integration, audio guides and a marketing site - all in just a handful of days
Actually useful in real life - I’ve been using ChowPilot to cook dinner myself!
Found a fun, brand-safe name with ChowPilot.com
Met both the audio and creative domain hackathon challenges

What I learned

Vibe coding tools really are powerful enough to create serious MVPs, fast. Coming into this I was sceptical and figured I'd quickly find myself tangled up in a spaghetti mess or blowing out context windows. I stand corrected.
The speed at which you can build in Bolt means you don't have to choose between building UI first or back end first - You can do the two in parallel and that makes for better products.
I thought talking to my computer would be a ridiculous experiment I'd only try once - turns out it's really productive. I may be hooked.

What's next for ChowPilot

✨ Custom voices per cuisine (imagine a French chef guiding you through Coq au Vin)
🎥 Add video recipe guides with humour, personality and culture. I really wanted to try this with Tavus but ran out of time.
🛒 Browser automation to auto-build shopping carts. Something I'm experimenting with locally.
📱 Build a mobile-first experience for easier use while cooking, possibly an app using Expo
🍽️ Deeper personalisation — allergens, protein preferences, portion sizes, blocked ingredients etc
🌏 Tighter integration with supermarkets to optimise for accuracy and cost