CuriosityPedia

Learn to be more curious. A beautiful on-demand mini-encyclopedia backed by reputed sources for exploring new rabbit holes.

Comment

Inspiration

I am a very visual learner, and I loved encyclopedias as a kid. Being a visual thinker has pros and cons, cons being struggle using today's chatbots becuase they are not good visual explainer they dont proactively think and search for the images that are freely available today on the interent to explain something. And i literally struggle to understand somethig when i have to read stuff(words are intemidating), you can call it a "learning disability", but I love to observe an image deeply and noticing tiny details which might interest a lot of people. Pros being I can truly appreciate the beautiful world around me.

I thought someone will figure it soon and make this visual communication capable chatbot, and that wait was about 3 years long now. I havent seen any product which makes the chatbot with great visual communication. And with this hackathon i saw an opportunity to show how the information consumption can become fun and exciting.

What it does

It gives a beautiful 8-10 pages of encyclopedia for a thing you want to learn. At the end of the encyclopedia, it asks questions about the images you saw. After answering the questions user can choose any question to understand in more depth and create encyclopedia for those topics.

User can see the journey they took into the rabbit holes when they explored the topics one by one,

How we built it

I used GPT 5.6 Sol Medium, and brainstormed the product like the User interface, then with the image generation capabilities I designed the UI of the application.

And then spent a lot of time coding(codexing) and deploying the website iteratively, I used sites for deploying the website, and used openai API as the brains of the application.

Challenges we ran into

TLDR Time management and token management.

Design was a major challenge image generation in codex is good, but still there is a lot of room for improvement, the images are not easily controllable, and not consistent between various screens of the application.

GPT5.6 Terra made me loose a lot of time as well, I started with GPT5.6 Sol which made things happen but when i came close to my quota limit, I had to shift to Terra and that forced me to get in the weeds of the code from time to time and would be blind to a lot of obvious errors or issues.

Another major challenge to build something that is simply educational and not a social-media or a chatbot was the vision of the product. It is difficult to explain how general purpose both the products are, and that when you are using certain software every day, it becomes hard to think outside of that framework and the northstar I followed was to make something that I would love to use.

Accomplishments that we're proud of

I am proud of the product that I have made it is close to what I imagined, but even more importantly I was able to build something like this by yself in 1 week, this gives me immense joy and confidence.

What we learned

Understanding model capabilities is very important, cost is an important aspect, but capabilities is more important because cheaper model just cant do something that frontier can.

I learnt that it is really important to plan well, pivots are not cheap in terms of time and token. Although pivot doesnt mean you failed to plan, sometimes you dont know what you want until you start building something and use it. Some ideas soud great in the head, and some simple ideas are still not implemented becasue they didnt sound so cool in the head.

What's next for CurisosityPedia

I think this is a great tool for the kids to learn new things, see things for the first time from the eyes of the best photographers. I want to create an agentic-editorial staff for this product, right now it is all done by 2 agents, but in future I would really like this to become this one stop soltuion for anyone wanting to learn about the beautiful world we live in and

I havent used frontier in the API keys as i cant afford that much experimentation right now, but I will improve this product with newer models, better visual interface for people like me.

Built With

Share this project:

Updates