Inspiration Zico was inspired by the idea of creating a lightweight AI assistant that feels natural and adaptive instead of generic. I wanted to explore how an AI could respond differently to greetings, goals, and random thoughts while maintaining useful conversational context in a simple interface. What it does Zico is an AI-powered assistant that responds intelligently to user input by maintaining conversational context and adapting its responses over time. It handles greetings, goal-oriented inputs, and reflective thoughts naturally, while keeping replies concise. Optional expandable explanations allow users to understand responses without cluttering the main chat. How we built it The project was built as a single-page web application using React, TypeScript, and Vite. It integrates GPT-5 through the Comet API to generate context-aware responses. Additional logic classifies user input to guide how the AI responds, and the UI focuses on clarity, responsiveness, and fast interaction. The app was deployed using Surgent. Challenges we ran into One major challenge was controlling AI behavior—making sure the assistant didn’t treat every input as a goal and handling conversational context correctly. Another challenge was preventing internal AI reasoning from being displayed while still allowing optional explanations. Time constraints also required careful scoping to ensure a working demo was completed. Accomplishments that we’re proud of Successfully integrated a production-grade AI model (GPT-5) within the hackathon timeframe Built a context-aware assistant that responds naturally to different types of input Designed a clean UI with expandable explanations instead of verbose responses Completed and deployed the project solo under time pressure What we learned This project deepened my understanding of AI interaction design, context management, and prompt control. I also learned how small UX decisions—like hiding internal reasoning or classifying inputs—can significantly improve how intelligent an AI feels to users. What’s next for Adaptive Suggestions for Daily Flow Next steps include improving long-term memory across sessions, adding clearer context reset controls, and expanding personalization so suggestions adapt more strongly to user habits over time.

Built With

Share this project:

Updates

Submission history