FluxMind: Simulate Your Week Before It Breaks

🔥 Inspiration

As students, we are constantly told to "manage time better" — but in reality, time is not the real problem. The real problem is uncertainty under overload.

Assignments take longer than expected, deadlines cluster unpredictably, and life events constantly interrupt carefully made plans. Traditional tools like calendars, Notion, or todo apps assume that time is deterministic — but student life is not.

We built Overload Radar to answer a different question:

Not “how do I finish everything?”, but “what should I cut before I break?”

🧠 What it does

Overload Radar is a cognitive load simulation engine for students.

Instead of generating fixed schedules, it:

Models tasks as uncertain time distributions
Simulates multiple future scenarios (Plan A / B / C)
Predicts stress trajectories under different decisions
Evaluates what should be cut, delayed, or preserved
Provides optional AI-generated voice support for overload moments

It transforms planning into decision simulation under uncertainty.

⚙️ How we built it

We designed a full-stack system combining:

Next.js (frontend + API routes)
Gemini API (Google AI Studio) for:
- task understanding
- probabilistic time estimation
- scenario reasoning
ElevenLabs API for:
- calm and emergency emotional voice support
Zustand for lightweight state management
TailwindCSS + shadcn/ui for UI system
Framer Motion for interaction dynamics

The architecture follows a strict separation:

AI models → reasoning only
Backend → orchestration layer
Frontend → deterministic visualization layer

No AI calls are made directly from the browser.

⚡ Key Innovation

Unlike traditional productivity tools:

1. Uncertainty-aware modeling

Tasks are not fixed durations.
They are modeled as ranges:

min time
max time
expected time
variance
confidence score

2. Scenario-based planning (A / B / C futures)

We do not output a single plan.

Instead, we simulate multiple futures:

Plan A: Stable execution (no extra events)
Plan B: Balanced execution (some life events included)
Plan C: High-pressure execution (deadline compression)

Each plan includes:

stress curve
feasibility score
what must be sacrificed

3. What-if simulation engine

Users can inject unexpected events:

“What if I attend a 4-hour NBA playoff game this week?”

The system simulates:

how workload shifts
which tasks get delayed or dropped
how stress changes over time

4. Emotional support layer

We added an optional AI voice layer using ElevenLabs:

calm advisory voice for normal guidance
emergency comfort voice for overload moments

This is explicitly opt-in and designed as AI-generated support, not human impersonation.

💥 Challenges

Designing a system that moves beyond “task planning” into decision simulation
Handling uncertainty in task duration without overfitting deterministic assumptions
Building meaningful stress modeling instead of simple scoring
Balancing product clarity with complex underlying simulation logic
Ensuring AI outputs remain structured, consistent, and explainable

📚 What we learned

Planning is not the real problem — decision-making under uncertainty is
Students do not need more tools — they need better tradeoff visibility
AI is most powerful when used as a simulation engine, not a chatbot
Emotional support systems must be designed with clear consent and boundaries