Eureka

Inspiration

The idea for Eureka was born from a simple, frustrating experience that we can all relate to: getting stuck on textbooks and PDF readings, and instead of clear, visual explanations, students are left with dense text and only their own imagination. Believing that visual learning is what students truly enjoy and prefer, we built Eureka, translating abstract STEM ideas into an intuitive animation when the student needs it.

What we built

Eureka is a Chrome extension with collapsible panels that lets you paste or highlight any STEM concept (math, physics, computer science, or chemistry) to generate two things: A 1-2 minute manim-style animation that visually explains the topic, and a set of 3-4 practice problems that are meant to reinforce the idea, complete with solutions and explanations.

Eureka combines a Chrome UI, with a Node.js backend that uses the Groq API to interpret the user's concept to generate an explanation summary, key points, an animation script, and the practice problems. Then, it triggers an AI video generation to turn the script into a short video, hosted online and returned as videoUrl in the JSON.

Eureka is unique in that it responds to the broader trend of passive learning platforms. Instead of sending learners to a new tab, Eureka lives inside the browser sidebar, allowing the user to use Eureka wherever they already study: Wikipedia, Khan Academy or even a scanned PDF. We followed our mission that learning should be on demand.

How we built it

Most, if not all of the frontend was done by prompting Figma Make, a tool we found to be very helpful in designing our product into a more polished look. The backend is simply just a Node.js server, with two endpoints /explain, and /practice.

Challenges we ran into

One of the biggest challenges we ran into is the Latency and reliability of AI video generation.

Generating a manim-style explainer video on the fly added noticeable delay, and the quality of the AI-generated videos would sometimes be subpar. Given the time constraint, we had to try our best to balance quality with speed.

Our other big challenge was integrating Groq with structure outputs

Getting the LLM to consistently return clean JSON-like structure for explanations and practice problems took iterative prompt engineering.

Accomplishments that we're proud of

Getting a working MVP up and running within a couple hours was an accomplishment in itself. Starting off with a completely different track and idea and transitioning into the education track in about 12 hours is something that our whole team is definitely proud of.

What we learned

The ability to debug multiple files and understand how the different layers of code are connected and act together is something we came in with minimal experience and left with far more of.

What's next for Eureka

We understand our limitations in our current AI-video generation, and we hope to integrate a true manim-style video renderer, that runs entirely in the browser, such that the animations can be generation directly from the JSON script inside the panel instead of external APIs. We also hope to expand this to not only students learning academic concepts in the STEM field, but also in the Fine Arts industry and even in the far future students looking for efficient interview prep.