Inspiration
ChatGPT is great! But switching tabs every now and then to understand topics is frustrating.
We get it. When we were reading a paper on Fuzzy KNN, we thought, “Okay, this can’t be that bad.” Two minutes in, six ChatGPT searches later, and we were still drowning in jargon. It felt less like reading a research paper and more like deciphering an ancient manuscript—except with way more acronyms.
Reading an article, a research paper, or even technical documentation shouldn’t feel like a scavenger hunt for definitions. We built a better way—a tool that works with you, right where you are. Our AI-powered research assistant extension brings instant explanations, contextual insights, and expert-level understanding without ever leaving the page.
No more tab-hopping. No more breaking your flow. Just smarter, faster learning.
The Vision
The way we consume information is evolving, but the tools we use to understand it haven’t kept up. Yet, until now, there hasn’t been a tool that truly integrates into the way we read and learn—one that is simple, intuitive, and always there when you need it.
Our vision is straightforward: Knowledge should be always at your fingertips. Whether you're a student, a researcher, or just someone curious about the world, knowledge should come to you—not the other way around. With features like real-time explanations, contextual awareness, and future support for graph breakdowns, flowchart insights, and visual learning, we’re making research and learning more accessible than ever. There are plenty of AI tools out there, but none that work this seamlessly, this user-friendly, and this well inside the content you’re already engaging with.
What it does
Instant Explanations – Highlight any text, and our AI will break it down for you in real time. Seamless Integration – Works on any webpage, research paper, or technical document with zero disruption to your reading flow. Adapts to Your Learning Style – Whether you need a quick definition or a detailed breakdown, it tailors explanations based on your setting. Understands Context – It doesn’t just pull dictionary definitions—it considers the surrounding content to give relevant explanations. Grows with You – As you refine your difficulty preference and add context, the AI refines its answers, making responses more precise over time. Eliminates Cognitive Overload – No more sifting through paragraphs of unrelated information—just the exact insight you need. Works Silently in the Background – It doesn’t clutter your screen or get in the way. The moment you need it, it’s there.
How we built it
Frontend (Popup UI & Toggle Control)
The extension's interface is clean and minimal, designed using HTML, CSS, and JavaScript. We implemented a toggle switch so users can enable or disable the extension easily. The popup also includes a "Context Aware" button, which sends the full page’s content to the AI for more relevant explanations. It also has a Details Control button which will allow you to choose how detailed you want your answer to be.Content Script (Injecting AI into Webpages)
The extension listens for text selections using JavaScript event listeners. When a user highlights a word or phrase, we inject a floating icon next to it, which acts as a trigger for explanations. Clicking the icon opens a dynamic response box containing the AI-generated explanation.Persistent AI Chat with Google’s Gemini Model
We integrated Google Generative AI (Gemini 1.5-Flash) to power real-time explanations. Instead of treating each query as a separate request, we maintain a persistent chat session so that explanations can build on previous responses, making the AI more contextually aware.Storage & State Management
We use Chrome Storage API to remember user preferences, such as whether the extension is enabled or disabled. This ensures that the experience is smooth and customized for each user without needing to reset preferences on every page reload.Performance Optimizations
To keep the extension lightweight and efficient:- We cache explanations to avoid redundant API calls.
- The response box disappears when clicking outside, keeping the page clutter-free.
- The AI only activates on demand, preventing unnecessary processing.
- We cache explanations to avoid redundant API calls.
Challenges we ran into
Engineering Challenges
Ensuring Seamless Integration
Many research tools disrupt the reading experience. We designed our UI to be unobtrusive—no pop-ups hijacking the screen, just a small icon that appears when needed.Fast Response Times
We had trouble getting Gemini to provide responses as quickly as it was generating them. Initially, the model would process the request in full before returning a response, leading to noticeable delays. To improve this, we learned about response streaming, which allows us to receive and display parts of the response as they are generated. This significantly reduced wait times, making interactions feel much more real-time and responsive.
Other Challenges
- User Experience & Simplicity
Making an AI tool that enhances learning without feeling intrusive required careful design. We constantly asked, Would this feel natural while reading? and refined interactions accordingly.
Accomplishments That We're Proud Of
A Real-World Solution That People Want
This isn’t just a cool project—we know people who need it. We’ve had people asking for early access because they see how it could fit into their workflow.Building This as a Two-Person Team
Every line of code, every design decision, every feature—you name it, we built it. As a two-person team, getting this far in such a short time is something we’re incredibly proud of.A User Experience That Stands Out
Research tools often feel clunky or get in the way. We truly believe our UI is the most user-friendly of any similar tool that exists in the internet right now. It’s seamless, intuitive, and actually helps rather than distracts.
What We Learned
Simple Solutions Can Be the Most Impactful
We didn’t reinvent the wheel—we just made accessing knowledge more intuitive. Sometimes, the best innovations come from removing friction rather than adding complexity.Balancing AI Power with User Experience
AI can do a lot, but throwing an entire research paper at it isn’t always the best approach. We learned how to optimize context extraction to make responses both relevant and efficient.Building for Real Users Changes Everything
Designing something that people actually want to use forces you to think differently. Getting early feedback helped us refine the experience, and we now understand the importance of iterating based on real needs, not just cool ideas.
What’s Next for Project Research Assistant
AI-Powered Image & Graph Explanations
Text is just one part of research—complex graphs, charts, and equations often hold key insights. We’re working on AI-driven visual analysis to break down these elements in a way that makes sense instantly.Multi-Source Context Awareness
Right now, we pull context from the page you're on. In the future, we aim to pull in related papers, articles, and sources to provide even deeper insights while you read.Enhanced Personalization
Not everyone learns the same way. We plan to refine difficulty levels further and introduce learning preferences, letting users get explanations that match their style—whether they prefer analogies, formal definitions, or step-by-step breakdowns.Offline Functionality
A truly seamless research tool shouldn’t always depend on an internet connection. We’re exploring ways to cache key concepts locally, so users can access explanations even when offline.Community-Driven Expansion
Some of the best features come from user feedback. We’ll continue gathering input from researchers, students, and curious minds to shape future updates, ensuring this tool grows with the people who use it most.
Log in or sign up for Devpost to join the conversation.