Inspiration

When I'm reading a textbook, I try to fully understand each concept before moving on to the next because I'm really bad at cramming, and not the most brilliant student either, so I've come to rely heavily on AI applications like ChatGPT to walk me through difficult material. So what I did was to scan the pages or paragraphs I didn't understand using my iPhone, extract the text, and paste it into ChatGPT to simplify it. But this came at a cost. The AI never had enough context to truly understand what the text was saying, and there were often other parts of the textbook connected to what I was trying to grasp that it simply had no access to. I couldn't upload an entire 300MB textbook to ChatGPT either ,it would take forever to process, and on the free plan, it might not even work at all.

I endured this workaround for a long time. But the day I decided to actually solve the problem was during a viva exam. I was asked a question and failed to answer it. My lecturer, out of grace, told me to go look it up in the textbook and come back with the answer. He gave me five minutes — nowhere near enough time to search through an entire book. So I turned to ChatGPT, but it kept confusing two similar terms, because I was unintentionally feeding it the wrong answer, and it just kept agreeing with me. Something in me knew this wasn't right. Eventually, one of my lecturers noticed me struggling and stepped in to help. That was the moment I realized enough was enough: I needed to build an AI that could actually read through my large textbooks and answer my questions properly — because the only reason my lecturer could help me in that moment was that he had already read the entire textbook himself.

I have to fix this.

What it does

On our website gd1.online users can upload textbooks as large as 400mb in seconds and use our ai feature to ask questions concerning any thing from the textbook in less than 15 secs , we also have features like my coach , that allows users to generate different styles of practice questions with questions and answers gotten from the files or textbook uploaded

How we built it

The first phase of the app was gotten from lovable, it was just like a foundation, from there I moved the folder to Codex to really bring the app to life. I used Claude to solve really complex problems i noticed Codex couldn't solve but one thing was that the frontend design was made to the peak with Codex 5.6, but the main features of the app like the speed to collect large files in seconds works by chunking it. The AI uses vectors to really give a deep dive into questions asked by users. I use both DeepSeek and OpenAI 4o for the response; DeepSeek does the heavy lifting, like looking through the textbook, while the response is given to OpenAI to reply in a more personalized way.

Challenges we ran into

The main challenges I experienced were looking for the AI model to search through the chunks without running down my credit and the most recent was the problem with scanned documents, which meant that we couldn't chunk it so we had to use OCR to extract text, but still, that isn't really effective and it can take large files up to 400 mb like a basic softcopy textbook

Accomplishments that we're proud of

Apart from seeing my problem being solved, my biggest accomplishment was seeing that my website could do what Claude, ChatGPT, Gemini, Grok, and DeepSeek couldn't do, it was a beautiful day for me, but that's still the beginning for what the future awaits me

What we learned

Don't sell features, don't sell the product, but understand the pain point of people, and show them there is a way out

What's next for G&D

Right now, I just turned the website to be specific for medical and law students, it will give me enough power to modify the website in a way that once it really satisfies my needs as a medical student, it will definitely satisfy someone else, plus i have a smaller market to concentrate on

Built With

Share this project:

Updates