This is a submission for Weekend Challenge: Generosity Edition
What I Built
Every charity app I've used starts with the same question: how much money can you give. If the answer is not much, you feel a bit guilty and close the tab.
That's the actual problem I wanted to solve. Most people don't skip giving because they're selfish, they skip it because the ask never fits what they actually have. Two hundred rupees feels too small to matter. An hour on a Saturday doesn't feel like "volunteering." An old laptop just sits in a drawer.
Spare flips the question. Instead of asking what you can donate, it asks what you have spare right now, whatever that is: some cash, a free evening, a skill, a language you speak, an object you don't use, access to something like a car or a rooftop. You type it in plain language (any language, any mix, doesn't matter), and Spare reflects it back to you as a little inventory, then gives you three specific things you can actually do with it this week. Not "go volunteer somewhere," but the actual organization, the actual first step, and a message already drafted for you to send.
The thing I cared about most while building this: most tools like this stop at "here's a place near you" and leave you to figure out the rest. Spare tries to close that last gap, because that's usually where good intentions die.
I also didn't want to just trust whatever an AI tells me is a real charity. So every result comes with a label telling you exactly how sure Spare is: confirmed live through search, hand checked by me beforehand, or "search this yourself" if neither of those worked out. I'd rather show you three honest results than five made up ones.
Demo
Live app: https://spareapp.ai.studio
Try the demo personas on the first screen if you don't want to type anything, they'll walk you through the whole flow in a few taps.
Code
GitHub repo: https://github.com/dhruvvvgg/spare
How I Built It
I built this entirely on my phone using Google AI Studio, so the whole thing had to work within that constraint from the start.
The matching works in three layers. First, Gemini does a live search grounded lookup based on what you typed and where you are, and only returns something if it can actually attach a real source link to it, nothing gets shown without a citation. If that comes back empty or thin, it falls back to a list of organizations I manually verified beforehand, checking every single link myself before it went into the app. And if even that doesn't have a good match for what you're offering, it hands you off to a live search on Idealist instead of pretending to have an answer.
I spent a good chunk of the weekend just verifying that fallback list by hand, going through Goonj, Robin Hood Army, eVidyaloka, ConnectFor, HelpAge India, Bhumi, and a few global ones like Translators without Borders and Be My Eyes, checking that each org still exists, still does what it says, and has a live link. It felt slower than just letting the model invent something, but for a project about charity, the last thing I wanted was to point someone toward a fake or defunct organization.
Once you pick one of the three results, a second Gemini call drafts an actual outreach message in your own tone, based on how you phrased your original input, so you're not stuck staring at a blank text box trying to figure out what to even say to a stranger or an NGO.
Prize Categories
Submitting for Best Use of Google AI. Gemini runs the entire pipeline here: pulling structured data out of free text, doing live grounded search for real local opportunities, and writing the final outreach message in your voice.
Top comments (0)