Deep Learning

Inspiration

In doing an online course earlier this year, I realised that learning is not easy. I tried to imagine what it would look like to build a tool that helps me manage my learning, by keeping track of what I have learnt so far, and therefore guiding me on the next thing to learn. I also came across some interesting ideas about thinking ("It takes two to think") and realised that some of my best thinking happens when I am forced to verbalise my thoughts. I therefore wanted to build a learning platform that makes this "thinking out loud" an integral feature of the learning experience. Overall, I felt like thinking and learning are experiences that would benefit from being immersive, and I set out to build a tool that allows you to engage all your senses fully in the learning process.

What it does

It allows students to organise their course material, then access it through a reader in the app. They can annotate the learning material while they are reading it (annotations are saved for future reference). They also have a canvas available to take more detailed notes and draw diagrams if needed along the way. Lastly, there is a voice tutor with whom they can converse, to understand concepts better when reading notes, or to be guided along in a Socratic way when working on a problem they can't quite solve.

How I built it

In a rush - I found out about the challenge way too late, and had 8 hours to build the project from scratch. I started by defining the key features and laying out the tech stack for Codex, and then asked Codex to start building the app while guiding me on what I need to do to get the deployment working (set up supabase, Vercel). Along the way, if features did not work as expected, I would feed this back to Codex to understand why, and to brainstorm alternatives e.g. using nutrient.io as a in-browser PDF reader and editor.

Challenges we ran into

Primarily the lack of time made it hard to build a more complete product - I have just built the interaction layer, and did not get a chance to implement more long term learning features I think would have made the app useful (for example, generating flashcards at the end of each session based on the student's key weaknesses noted during the lesson). Also, I did not have access to the $100 Codex credits because I signed up too late, so I was stuck with the free tier and limited usage. I had to eventually supplement this with my own subscription to an external coding agent.

Accomplishments that we're proud of

Getting the entire stack (text, free form annotation, canvas, voice) working together.

What we learned

Being clear on features and tech stack makes the development process really fast.

What's next for Deep Learning

I would like to build up the long term learning features I have mentioned, and get it in the hands of users to get feedback on whether it really is effective in maintaining engagement in learning. Also, enhance the PDF reading and editing experience in the browser.