Gemini Assumption Breaker

Inspiration

Many poor decisions are not caused by a lack of information, but by unchecked assumptions.
People often act on beliefs without questioning whether those assumptions are valid, fragile, or outdated.
I wanted to build a tool that helps users think better, not just get answers.

What it does

Gemini Assumption Breaker is an AI decision-support tool.
The user enters a belief, plan, or assumption (for example: “Data Science is safer than Computer Science in 2026”).

The app:

Identifies the hidden assumptions behind the belief
Evaluates how strong or fragile those assumptions are
Explores what happens if those assumptions fail

Instead of giving a single answer, the tool:

Extracts underlying assumptions
Rates their strength
Simulates counterfactual outcomes
Reframes the belief with balanced insights

How I built it

The application is built using a simple web interface with the Gemini API as the reasoning engine.
Gemini is used for structured reasoning, assumption extraction, and counterfactual analysis — not simple text generation.

Due to API quota limitations during the hackathon, the app includes a demo mode that uses representative sample Gemini outputs while preserving the exact reasoning structure.

Challenges I ran into

The biggest challenge was balancing deep reasoning with a clear and simple user experience.
Another challenge was working within limited API quotas, which required designing a fallback demo mode that still accurately demonstrates Gemini’s reasoning capabilities.

Accomplishments that I'm proud of

Designed a novel use case for Gemini focused on reasoning, not chat
Built a transparent demo mode without hiding limitations
Created a tool that encourages critical thinking, not quick answers

What I learned

I learned how to design AI systems that support decision quality, not just information delivery.
I also gained experience structuring prompts for multi-step reasoning and counterfactual analysis using Gemini.