Reinforcement Quick Demo

## Inspiration

I'm someone who has been exploring the breadth and depth of Claude Code's capabilities. I've been building with it since December, but I've learned a great deal from it and thoroughly enjoyed using it to build helpful tools or just for fun.

I was inspired to build Reinforcement because this is actually a tool I've been wanting to make for
myself, as a student. I've really appreciated having a tutor in Anthropic's Claude. I'll download files
to my laptop and upload them to Claude to ask for help in understanding challenging topics from Data
Structures and Algorithms, as well as Discrete Math. What I end up doing a lot of is screenshotting bits and pieces of a slide deck and asking specific questions relating to those screenshots. This is what inspired my draw, or lasso feature, within Reinforcement's Content view. A user is able to draw a circle around the portion of an uploaded slide deck and chat right there in the same place about that snippet. Clean, quick, easy, and clear.

Another source of inspiration is one I'm sure many can appreciate: textbook summarizing! This is nothing
all that extravagant; anyone can upload a text to AI or ask for general summaries. But Reinforcement goes a step beyond and additionally provides a Quiz Me functionality, where you really make sure you are getting that text content — that you may or may not have read fully — down.

## What it does

Reinforcement is a Western-themed AI study platform. You upload your course materials — PDFs, images, or text files — and interact with them through Obadiah Wickett, an AI sheriff companion who helps you learn, review, and test your understanding. It features a content viewer with a lasso tool for circling parts of slides to ask about, configurable quiz generation from your materials, and a one-pager study guide generator — all with a persistent AI chat assistant alongside.

## How we built it

I built this project entirely using Claude Code. This made this fun project a day's work, a piece of cake, rather than weeks of coding and head-scratching. The stack is Next.js with TypeScript, Tailwind CSS, Neon Postgres with Drizzle ORM, and the Claude API for all AI features including streaming chat, quiz generation, one-pager generation, and vision-based content extraction.

## Challenges we ran into

Yes, I say it was a piece of cake, but I would be lying if I told you I didn't come with challenges. The challenges I faced were pretty common to challenges I've already faced while using Claude Code. It mainly struggles with content display, trivial though that may sound. It is sometimes the most difficult part for Claude, as it will struggle with interpreting what exactly you want and how to execute it. It takes careful prompting, attention to detail, and a good helping of patience. It really is amazing how well it does with even these tasks, though.

The time limitation was also a challenge, as I have more ideas and hopes for this product, which I know
would make it even more valuable and irresistible.

## Accomplishments that we're proud of

I also wanted my product to stand out from other tools like it, which I think it does. I'd already prefer my tool to several others I've seen online. The lasso feature in particular — being able to circle any part of a slide and immediately chat about it, is something I haven't seen in other study tools. It was an idea I was excited to bring to life. And surprised by how it didn't actually take too long, with the help with Claude.

## What we learned

I learned more about how to leverage my tools to build something new end-to-end. Building with Claude Code since December has taught me a great deal, and this hackathon was a chance to put all of that experience into a single focused sprint.

## What's next for Reinforcement

I have some more ideas and hopes for this product that I didn't include in my, not so minimal, MVP. I want to add multi-class session support for instance, so you can organize materials by course. Mobile optimization so it works just as well on a phone, and much more. Simply improving for more smooth usage.

Видео Reinforcement Quick Demo канала Jasmine Erkel