Loops HouseSecret Sauce

NEAR HACKATHON
Links
Team
1 member- ADOwner
Adam Timur Aslan
Overview
Secret Sauce won 1st place among 10 teams at the Build City NYC NEAR Private AI Hackathon. The organizers called it "Commercialized refining of local agents". Our ambition: turn a business's hard-earned know-how into its own private, continuously improvable AI agent. Restaurants are our first market.
A gate server on the owner's Mac connects local files and restaurant books to controlled AI tools. Questions and selected tool results go to NEAR AI Cloud for confidential inference. The agent proposes memory notes and house-playbook lessons; the owner approves what gets saved. A local Weight Lab lets the model experiment with small weight edits and request an undo. This is a refinement loop with the business owner in control.
Our narrated sample-data walkthrough shows the restaurant use case and the vision for the next expansion: NEAR Intents could handle vendor payments, staking could fund inference, and an on-chain diploma could recognize a trained agent. The opportunity is bigger than a chatbot: business-specific intelligence with a path to autonomous operations and real-world commerce on NEAR.
Built by Luis Escobar, Bondan Herumurti, Sahil Pardasani and Adam Timur Aslan. This submission extends the winning project with a new narrated video and illustrated README. The restaurant walkthrough uses simulated Odoo data. Original project: https://github.com/Escobar-Luis/secret-sauce. Source code remains MIT-licensed; new video and narration copyright 2026 Adam Timur Aslan.