DualMind

MainScreen
TwoChoices
FinalResult

Inspiration

We were inspired by the idea that personality is not defined by what people say about themselves, but by the choices they make under pressure. Traditional personality quizzes feel static and predictable. We wanted to build something dynamic — an AI system that analyzes behavioural patterns instead of fixed answers. The concept of internal duality — rational vs impulsive, long-term vs short-term — became the foundation of DualMind.

What it does

DualMind is an AI-powered moral decision engine.

Users face a series of dilemmas and choose between two contrasting responses:

Rational, ethical, long-term thinking
Self-interested, impulsive, short-term thinking

After multiple decisions, AI analyzes the user’s behavioural pattern and generates a structured psychological profile including:

Persona name
Dominant trait
Shadow trait
Literary parallel
Personalized psychological insight

Instead of scoring points, DualMind interprets decision patterns.

How we built it

We built DualMind using:

Next.js (App Router) for the full-stack framework
TypeScript for type safety
Gemini 2.5 Flash for AI-driven analysis
A secure server-side API route to protect the API key
Basic rate limiting and input validation to prevent abuse

The system collects user decisions, constructs a structured AI prompt, and enforces JSON-only responses for consistent, scalable output.

Challenges we ran into

Enforcing structured JSON responses from AI consistently
Handling malformed outputs safely
Preventing API abuse while keeping the experience seamless
Designing prompts that produce psychologically meaningful insights instead of generic advice
Balancing immersive narrative tone with structured technical output

Accomplishments that we're proud of

Successfully built a behaviour-driven AI personality engine
Implemented secure server-side AI calls with protected API keys
Created structured JSON outputs for scalability
Designed a clean decision-to-analysis flow
Turned a philosophical concept into an interactive AI product

What we learned

AI prompting requires iteration to achieve depth and consistency
Behavioural framing produces more meaningful insights than static scoring
Structured output enforcement is critical for reliability
Even simple rate limiting significantly improves production readiness
Strong product naming and brand changes how the project feels instantly