Results

PaintPal is a friendly web based drawing app meant to create artwork that can be shown in Spectacles. Using Subabase on the backend and Figma Make for front end, Paint Pal makes it easy for anyone to create digital artwork. Future versions will integrate Gen AI to create photorealistic versions of sketches. This way anyone can be an artist!

A mobile app that records CrossFit workout in seconds and instantly shows what you’re improving, and what to train next. (test user: test@gmail.com / pw: test) People doing high-intensity, functional workouts often track results in scattered notes or spreadsheets, so progress is hard to see and training choices are guesswork. This app lets you log a workout in a few taps and automatically analyzes your performance across 10 core fitness abilities (like endurance, strength, speed, and coordination). You get clear scores, simple visuals, and bite-size tips so every session builds smarter toward your goals. Made by using Figma Make & Supabase

Test user: email: test@test.com pass: Test123 PromptStudio IDE is a comprehensive prompt engineering workspace that helps you create, test, and optimize AI prompts with precision. The application features a powerful editor with variable extraction, version control to track every iteration of your prompts, and an integrated testing system that runs prompts with different variables while capturing detailed metrics like token usage, costs, and response times. The IDE includes AI-powered improvement suggestions that analyze your prompts and recommend optimizations, which you can apply with a single click. Each test result is automatically compared against previous runs, showing performance improvements or regressions with color-coded indicators. The version history system lets you restore previous iterations or compare changes using a built-in diff viewer. With markdown rendering for responses, a variables panel for managing dynamic content, and Supabase integration for persistent storage, PromptStudio provides a complete environment for professional prompt engineering and iteration.

Generate ERD Diagrams for your applications just by chatting with AI. Describe your app ("a todo app with users and comments") get an interactive ERD visualization, and deploy the SQL directly to your Supabase project with one click.

Video content increases conversion rates by 80% (https://firework.com/blog/how-video-content-boosts-conversion-rates), but most stores can't afford professional product videos. Shooting new videos for existing inventory will also cause a lot of money and resources. Our solution: A plug-and-play backend pipeline that automatically transforms existing product images into AI-generated videos using Higgsfield's Image2Video API. Store owners simply integrate a plugin into their backend—their static image catalogs instantly become engaging video experiences. Built with Next.js (faster api servers may exist, but chose next.js due to familiarity for now), Inngest for async processing, and Supabase Storage for fast delivery. The system handles thousands of products at scale through webhook-based video generation. The result: Any e-commerce store, from small businesses to large retailers, can now offer immersive video shopping experiences without cameras, studios, or production costs. Integration is easy, the capability can be served using a node library, javascript library, or even Wordpress Plugin. We built a Demo via Chrome extension to show how we can transform any ecommerce store with static images into a video powered one.

Chamba.ai is an autonomous sourcing agent for manufacturers. Upload your supplier invoices, and our AI Agents scans hundreds of vendors in real time to find better prices for the exact SKUs you buy. Chamba first understands your cost structure and them sources the best deals for you to optimize your bottom line.

Blueprinter is an AI-driven context engineering engine that converts one-line ideas into complete, style-aware, and contextually adaptive development blueprints. It combines GPT-5, FastAPI, LangGraph, Supabase, and a VectorDB to build and evolve a deep understanding of your coding DNA—learning your syntax preferences, architectural habits, and design intent. Every generated plan is context-linked: editable descriptions update dynamically, and “Ask Copilot” enables fine-tuned refinements grounded in your historical codebase. The frontend (React + TypeScript + Tailwind) supports live plan visualization and real-time collaboration, while secure backend orchestration uses HMAC signing, RLS policies, and vector-based embeddings for contextual recall. Blueprinter doesn’t just plan code—it engineers contextual continuity, acting as an adaptive thought-partner that remembers how you build and amplifies how you think.

Record matches, analyze performance, and compete with fellow tennis players in San Francisco. Your complete tennis tracking solution for the ultimate game improvement.

iOS mobile app for tracking pets records: User creates account and creates family, invite members. Then adding pets. After that we can track and record everything related to pet health: Already done: 1. Feeding 1.1. Add food ( wet, dry, treat), calories content 1.2 Set daily calorie goal 1.3. Record feeding ( autocalculate calories towards daily goal) 2. Medication 2.1 Create medication plan, start\end date, how many times a day, type of medication 2.2 On main dashboard you can quickly record administration of medicine 2.3 Showing if still need to administer or done for today 3. Health Journal 3.1 Add any health event, categorize ( select from existing previously added, or automatically create new category) 3.2 Vector semantic search using Supabase AI. For example:
when was first asthma attack? How many times vomit in last month ? ToDo:
4. Notifications 4.1 iOS widgets 4.2. Live Activity

Vortal is a plug-and-play client dashboard for AI Voice companies. Instead of building custom portals for every client, Vortal automatically syncs with your Vapi (and soon other) agents and gives each business a branded dashboard with call analytics, transcripts, and usage data.

I built an app that helps people make peace with their past and know what to expect of/look forward to their future. It does this by helping them view their past, present, and future via an integrated view of easter spirituality/psychology lenses, all tied together. It ties all of these different methods together to better understand themselves. They shouldn’t need to understand what these different lenses are, but should be able to understand their lives better.

Every day, companies and support teams receive floods of emails (support tickets, bug reports, partnership requests) all landing in the same shared inbox. Manually reading and forwarding each message wastes time, delays responses, and often sends emails to the wrong person. MailSplit automates this process with an intelligent, AI-powered routing system built on Resend, Supabase, and Retrieval-Augmented Generation (RAG). Incoming emails are first captured through Resend, ensuring reliable ingestion and webhook handling. Next, MailSplit’s AI triage system uses NLP to analyze each message’s intent, topic, and urgency: distinguishing between issues like “billing error,” “technical bug,” or “feature request.” Then, Supabase takes center stage. We use it to store, manage, and query organizational documents such as internal wikis, team roles, and historical ticket data. They serve as the foundation for our RAG context. When a new email arrives, the RAG engine retrieves relevant context from Supabase, helping the model understand who is best suited to handle the message based on expertise, history, and workload. For instance, if an email mentions “API timeout,” MailSplit consults Supabase to find similar past issues and routes it directly to the engineer who resolved them. - Smart, context-aware routing using RAG and Supabase. - Reduced response time and human error. - Scalable across teams and departments. With MailSplit, companies turn bloated inboxes into streamlined communication. We ensure every email reaches the right person.

Peyote Code Context MCP Server Peyote is a Model Context Protocol (MCP) server that brings AI-powered semantic code understanding to tools like Cursor and GitHub Copilot. It ingests any repository, generates embeddings using OpenAI, and leverages Retrieval-Augmented Generation (RAG) to provide developers with contextually relevant code snippets and insights. With Peyote, AI coding assistants gain real awareness of an organization’s codebase — for example, when a backend developer adds a new API, a frontend engineer can simply ask Cursor how to integrate it, and Peyote delivers the precise implementation details. Use of Supabase Peyote uses Supabase as the backbone of its data and context infrastructure. All embeddings are stored and indexed using pgvector for fast semantic search, enabling efficient retrieval during AI-assisted coding. Supabase Edge Functions handle asynchronous workflows — including repository ingestion and status updates — while Resend is triggered through these functions to notify users once their repositories are processed and ready to use. This combination ensures smooth orchestration, real-time performance, and a reliable developer experience. Use of Partner Tools To ensure observability and polish, Peyote integrates DataDog, Resend, and Figma Make. DataDog tracks custom metrics like ingestion duration, embedding performance, and repository health for real-time visibility into system performance. Resend delivers branded email notifications that keep users informed throughout ingestion. The UI, designed entirely in Figma Make, visualizes these insights through a responsive dashboard that reflects live Supabase and DataDog data. Together, these integrations make Peyote a complete, production-ready platform that bridges powerful backend intelligence with an intuitive, developer-friendly interface.

A personalized women's health tracking application that integrates with Ultrahuman devices to provide AI-powered wellness forecasts. The app uses Supabase for authentication and database, with AI forecasting via Supabase Edge Functions. This app also takes in the user's food intake for the day and calculates the amount of protein coming from the food. The AI insights at the top is to help the user be able to make better health choices to live a healthier life style. This app will help women in their cycle tracking and will help them make smarter choices in the weeks leading up to their cycle to have a healthier and easier cycle.

More co-working groups in the world means more building and connecting. This is a super simple prototype app for helping people discover, join, and create co-working groups. This was a great opportunity to play with Supabase and meet awesome people. Thank you!

We’re building a wearable-AI photo app that lets you see the world through imagination, real time. 
Using Snap’s Spectacles, users capture real-life moments hands-free, send them to Gemini, which restyles them through AI into our own personal “World.”
 Current AI tools are after-the-fact — you create, then upload, then wait. Snap’s Spectacles already bridge physical + digital. Gemini can provide on-device or cloud AI vision + styling. There’s no platform yet that lets you see the world as imagination in motion. With the support of Supabase, you can step into your friends’ worlds in our app, compare how everyone reinterprets the same city through their AI style. We want to make AI part of how we see the world, not just how we edit it afterward.

【Description】Struggling to recall details from past events and LinkedIn connections? Our AI relationship manager solves this. Effortlessly record conversations on our mobile app. Our Chrome extension then displays key insights, conversation history, and future collaboration value directly on a person's LinkedIn profile. Get smart follow-up suggestions and instantly find people by asking the AI about your past discussions. Turn forgotten contacts into valuable opportunities. 【Note to Judges】: Since the official Chrome Extension is pending store review and cannot be published within a day, we are demonstrating its full functionality using a local host. Please refer to our demo video for a complete walkthrough.

PaintPal is a friendly web based drawing app meant to create artwork that can be shown in Spectacles. Using Subabase on the backend and Figma Make for front end, Paint Pal makes it easy for anyone to create digital artwork. Future versions will integrate Gen AI to create photorealistic versions of sketches. This way anyone can be an artist!

Chamba.ai is an autonomous sourcing agent for manufacturers. Upload your supplier invoices, and our AI Agents scans hundreds of vendors in real time to find better prices for the exact SKUs you buy. Chamba first understands your cost structure and them sources the best deals for you to optimize your bottom line.

Project Description Career Tree App is an AI-powered career planning and guidance platform built with Next.js 15, leveraging Claude Sonnet 4.5 to help users navigate their career paths intelligently. Core Functionality The application provides a dynamic, multi-level career exploration experience: Resume Analysis: Users upload their resume/LinkedIn profile, which gets analyzed to extract current job title, skills, experience, and education. Personalized Path Generation: Based on the user's background, interests, and goals (salary, timeline, hobbies), the AI generates three realistic career paths at each level. Progressive Career Tree: Users can select any path and receive three more advanced career options building on that choice, creating an expanding tree of possibilities. Actionable Roadmaps: For any selected career destination, the app generates 8-12 detailed, sequenced steps with specific resources (courses, certifications, projects) to achieve that goal. Interactive Chat: Users can ask questions and get career advice within the context of their profile and generated paths. Tech Stack Frontend: Next.js 15.2 (App Router), React 19, TailwindCSS, Radix UI, Framer Motion AI/LLM: Vercel AI SDK 5, Claude Sonnet 4.5 (via Anthropic), structured outputs with Zod validation Backend: Next.js API routes, Server Actions Database: Supabase (PostgreSQL with tables for sessions, career paths, steps, and conversation history) Email: Resend for notifications (plan finalization)

Browserstack but better. Simple and intuitive A/B testing using agentic, parallel AI workflows that tests features from your PR's or website. A/B testing with different UI flows has been notorious, with companies spending thousands of dollars to outsource the issue. We offer a simple and straightforward solution where AI agents can run in parallel and accomplish the same testing that companies dreaded.

Aurora is the friend who actually pays attention. She sees what you see, remembers what you said three hours ago when you were delirious from caffeine, and won't judge you for eating cold chicken at 4 PM during a hackathon. Need help? She's got you. Want her to roast your code? She can do that too. Unlike your real friends, she'll never ghost you.

Revisit the moments that mattered. Organise highlights captured from across the web, polish notes, and stay in flow as you move between stories.

alacard: The IKEA Assembly Layer for AI Models Transform any of 2.1M+ Hugging Face models into runnable Jupyter notebooks in minutes using a card-based recipe builder. Alacard solves the AI model adoption crisis by making it trivially easy to try, learn, and deploy any model through interactive, educational notebooks. Users mix Model Cards, Prompt Packs, Topic Cards, Difficulty Cards, and UI Component Cards to create customized "recipes" that generate working code examples tailored to their skill level and use case. The Problem: When new AI models launch, teams spend hours wrestling with bad documentation, failed setup attempts, and manual "vibe checking." Most of the 2.1M models on Hugging Face never get used because people don't know how to use them. The Solution: A card-based interface that generates runnable notebooks customized by your recipe. Mix cards → hit "Generate" → get a working .ipynb file with examples tailored to your topic, difficulty level, and UI preferences. Key Features: Shuffle: Remove analysis paralysis with "deal me a hand" button Generate: Customized notebooks based on your card combination Share: Every notebook gets a shareable link with community features Remix: Fork any shared notebook, modify the recipe, regenerate Trending: Community curation through usage metrics Technical Stack: Next.js frontend, FastAPI + Celery backend, Supabase database, WebSocket real-time updates, Hugging Face API integration. Impact: Reduces model evaluation time from days to minutes, democratizes access to open-source AI, and creates a thriving ecosystem of community-created notebooks that make any model approachable.

Autonomous bio-sonification transforming biological sequences into dynamic, intelligent musical compositions

I built an app that helps people make peace with their past and know what to expect of/look forward to their future. It does this by helping them view their past, present, and future via an integrated view of easter spirituality/psychology lenses, all tied together. It ties all of these different methods together to better understand themselves. They shouldn’t need to understand what these different lenses are, but should be able to understand their lives better.

We have built a real-time shopping assistant for diabetics that helps users make healthier choices while shopping. By wearing Snapchat Spectacles, a user can simply pinch while looking at a food product and receive an instant response with guidance on whether it’s a suitable option. For example your looking at can of full sugar coke it will say no this is not good for your health and give dietary reason why Disclaimer!!! the snapchat spectacles wouldn't let us record a demo, we have flagged this with the snapchat team and they said that it was on them as its a limitation of the dev kit.

AI-powered WhatsApp assistant that lets remote teams and individuals manage their finances through natural conversation. Stablecoin-first with Open Finance rails. Enhanced by AI.

Project Description Career Tree App is an AI-powered career planning and guidance platform built with Next.js 15, leveraging Claude Sonnet 4.5 to help users navigate their career paths intelligently. Core Functionality The application provides a dynamic, multi-level career exploration experience: Resume Analysis: Users upload their resume/LinkedIn profile, which gets analyzed to extract current job title, skills, experience, and education. Personalized Path Generation: Based on the user's background, interests, and goals (salary, timeline, hobbies), the AI generates three realistic career paths at each level. Progressive Career Tree: Users can select any path and receive three more advanced career options building on that choice, creating an expanding tree of possibilities. Actionable Roadmaps: For any selected career destination, the app generates 8-12 detailed, sequenced steps with specific resources (courses, certifications, projects) to achieve that goal. Interactive Chat: Users can ask questions and get career advice within the context of their profile and generated paths. Tech Stack Frontend: Next.js 15.2 (App Router), React 19, TailwindCSS, Radix UI, Framer Motion AI/LLM: Vercel AI SDK 5, Claude Sonnet 4.5 (via Anthropic), structured outputs with Zod validation Backend: Next.js API routes, Server Actions Database: Supabase (PostgreSQL with tables for sessions, career paths, steps, and conversation history) Email: Resend for notifications (plan finalization)

Kill AWS zombie costs in 10 minutes.