Inspiration
We've all experienced the feeling of studying for hours, thinking we understood everything, and then realizing during an exam that we didn't understand a concept as well as we thought.
We realized that one of the biggest challenges in learning isn't always finding information. It's recognizing what you don't know.
Most study tools help students memorize information or provide answers when asked. But what if students don't even realize they're misunderstanding something?
That's where Unlearn comes in. Rather than simply providing answers, Unlearn identifies potential gaps in students' understanding, helps uncover misconceptions, and provides personalized guidance to address them.
By making individualized academic support accessible anytime and anywhere, students can spend less time repeatedly reviewing concepts they already understand, avoid unnecessary trips to tutoring appointments, and focus their efforts on the material they actually struggle with.
Ultimately, we want to help students become independent learners who can recognize their own knowledge gaps, understand their mistakes, and take ownership of their education.
That's what inspired Unlearn: an AI-powered study companion that helps students uncover knowledge gaps, question their assumptions, and build genuine understanding.
What it does
Unlearn transforms studying into an interactive conversation where students explain what they know instead of simply asking AI for answers.
Powered by Google's Gemini API, Unlearn:
Asks adaptive follow-up questions to investigate incomplete explanations and potential misconceptions without immediately revealing the answer.
Tracks understanding by identifying concepts students have demonstrated knowledge of and areas that need further review.
Organizes study sessions into separate subject channels, allowing students to revisit their conversations without losing context.
Provides personalized feedback explaining why a concept may need more attention.
Unlike a general-purpose AI chatbot, Unlearn is designed with tutoring instructions that encourage students to reason through problems independently, without repeatedly having to remind the AI not to spoil answers.
How we built it
We built Unlearn using React, TypeScript, Vite, and Tailwind CSS for the frontend, with a Node.js and Express backend.
We integrated Google's Gemini API using the official @google/genai SDK to power adaptive tutoring and concept assessments.
We designed a custom tutoring prompt that encourages Gemini to ask targeted questions, challenge students' reasoning, and identify potential misconceptions rather than immediately providing solutions.
Using Gemini's structured JSON responses, we separated conversational replies from concept assessments. This allows the chat interface and understanding tracker to use the same AI interaction.
We also implemented browser localStorage to preserve study channels, chat histories, and understanding progress without requiring a database for our initial prototype.
Our interface was designed around an organized study workspace, featuring collapsible sidebars for subject navigation and understanding tracking, with the AI conversation at the center.
Challenges we ran into
Some challenges we ran into were figuring out how to translate our ideas into an actual working project. We had a lot of big ideas for Unlearn but we realized we would have to break it down into a smaller more managable project. We knew how a certain feature would look but actually implementing it into code was difficult.
There was also a lot of trial and error throughout the development process. Even features that seemed simple in theory required research, testing, and troubleshooting to get them working the way we envisioned. We also had to consider how individual components would fit together and how changes to one part of the interface might affect the overall layout and user experience.
Another challenge was balancing our original vision for Unlearn with the limited time we had during the hackathon. We had to prioritize the features that best represented our idea while making sure we could deliver a functional prototype.
Accomplishments that we're proud of
One of the things we're most proud of is creating something that comes from an experience we all understand as students. We all have had moments where we felt like we knew something only to find out we actually had gaps in our knowledge. This helped inspire Unlearn and we are proud we were able to turn a shared experience into a project that could help other students facing the same struggle.
We're also proud of how we approached the problem differently from traditional study tools. Instead of focusing on memorizing information or getting the right answer, we wanted to create a space where students could question what they know, recognize their misconceptions, and build a stronger understanding of concepts.
What we learned
Through Unlearn, we all gained hands-on experience with different aspects of web development. For instance, these include full-stack development, REST APIs, TypeScript, React, and Google's Gemini API.
We learned how to design AI behavior through system instructions, use structured outputs to turn AI-generated responses into usable application data, and manage persistent study sessions.
We also learned the importance of clear communication, modular code, and collaboration when developing a project under a tight deadline.
Most importantly, we learned that AI can be more valuable in education when it helps students think critically rather than simply giving them answers.
What's next for Unlearn
We envision expanding Unlearn with personalized review boards, targeted practice problems, PDF and lecture-note uploads, voice-based tutoring, and long-term learning analytics.
We also plan to introduce a database so students can securely access their learning history across devices.
Our goal is to make Unlearn a study companion that helps students develop lasting understanding and confidence in their knowledge.
Because learning isn't just about knowing the right answer. It's about knowing why.
Built With
- express.js
- google-gemini-api
- google-genai-sdk
- html
- javascript
- localstorage
- node.js
- react
- tailwind-css
- typescript
- vite