An interactive web simulation that teaches Secondary 3 students how an agentic AI plans and acts. I built it during my internship at the Ministry of Education (MOE), Educational Technology Division, Singapore (Jun – Aug 2026).
Students follow one scenario, an AI travel agent planning a two-week family holiday, from two sides.
Phase 1: User mode (you are the human). Chat with an AI travel app called Wanderly. Type requests, decide which permissions to grant (calendar, booking receipts), and watch the agent draft an itinerary, run into a problem and confirm bookings.
Phase 2: AI agent mode (you are the AI). Step inside the agent and make its decisions, one stage at a time:
- Constraints: pick out the hard facts in the user's request.
- Order: put the steps in a sensible order: understand, gather info, plan, then act.
- Plan: draft the sequence of steps the agent will take.
- Tools: match each action to the right tool, such as flights, hotels, calendar or web search.
- Memory and judgement: use what the agent knows about the family, and decide what to do when flights sell out.
Each stage gives instant feedback explaining why an answer is right or wrong, and a "thinking terminal" shows the agent's reasoning as it works.
Students use AI tools every day but rarely see how those tools decide what to do. AI Quests makes that reasoning visible, and shows why an agent needs clear goals, the right tools and a human who stays in charge. The design was refined over several rounds of user feedback.
React · Vite · Tailwind CSS · deployed on Vercel
npm install
npm run dev