CornerHarvest / core duo
CornerHarvest connects surplus from urban farms with corner stores in neighborhoods short of fresh food.
Major suppliers skip corner stores because their orders are small, while community gardens throw away surplus because they have no sales channel.
How it works
- A grower uploads a produce photo. Gemma 4 checks its quality and writes the listing.
- A forecast shows which neighborhoods will run short, and on which days.
- The app ranks nearby stores and plans a delivery route.
- The store bargains in plain language with the seller’s agent.
- Both agents record ratings on the Solana Agent Registry, so trust carries over to the next deal.
Built to be trustworthy: the grower’s minimum price is never sent to the AI, and fixed rules set every price, so a message can’t talk the agent below it.
Current status: the full flow runs in the app with fictional sample data. The Gemma, Snowflake, and Solana integrations are written. Gemma is tested against a local stand-in for the Gemini API, and the live Snowflake and Solana connections need account setup.
Built with
next.jsreacttypescriptTailwind CSSGemini API
Prize Categories Entered
- Submitted
Best Use of Gemma 4
- Submitted

Best Open-Source AI Project