Juiz

Inspiration

University life is a constant stream of deadlines, class schedules, expenses, files, and messages. The problem is not a lack of apps—it is the mental load of repeatedly deciding what to remember and what to do next. We built Juiz to make campus life feel lighter.

What it does

Juiz is a personal AI secretary for students. It turns everyday input into an organized, actionable view of campus life:

• Chat naturally to capture tasks, deadlines, plans, and expenses.
• Extract structured information from a single message and review it before saving.
• Keep calendar items, reminders, and Google Calendar synchronization in one place.
• Track spending and preserve important study materials in a searchable document vault.
• Surface the next useful action without forcing students to manage several disconnected apps.

The core principle is user control: Juiz asks for confirmation before writing important information.

How we built it

Juiz is a responsive web app and PWA built with Next.js, React, TypeScript, and Tailwind CSS. Supabase provides authentication and persistent data.

GPT-5.6 powers the intelligent experience inside Juiz. It understands natural-language requests, identifies multiple intents in one message, extracts structured events and expenses, and generates contextual assistance. In the demo, one Japanese message becomes both a proposed research meeting and a ¥2,400 expense, followed by a confirmation step before the event is saved and synchronized.

We used Codex throughout the build to inspect the codebase, plan features, implement and debug the Next.js and Supabase flows, verify integrations, and iterate on the user experience. Codex helped turn product decisions into working code while keeping the interface coherent across chat, calendar, money, and document views.

Challenges we ran into

The hardest part was making AI output useful without making it feel opaque. We designed an extraction-and-review flow so unstructured text becomes proposed structured data that the student can verify. We also had to keep the experience fast and coherent across several connected views while handling calendar synchronization carefully.

Accomplishments that we're proud of

• One natural-language message can create multiple structured suggestions.
• Important writes use a clear confirmation step.
• Calendar synchronization is verified after registration.
• The public landing page and the authenticated product form one cohesive experience.
• The demo shows a complete working flow rather than a static prototype.

What we learned

A helpful student assistant is not defined by how much it can say; it is defined by how reliably it reduces the next decision. Building Juiz taught us to pair model capabilities with clear UI feedback, confirmation points, and structured data rather than relying on a chat interface alone.

What's next

We plan to improve proactive planning, deepen document understanding, expand integrations, and make Juiz an even more dependable companion throughout a student's semester.

Built With

• openai-api