NOTENEST

The main body which displays your study desk.
The left sidebar
Reminders sticky note: To add any reminders, so that students/users can track what to do.
Pdf is given to the AI to read and your AI is ready to help you out.
With summaries of notes to highlighting important parts, your AI assistant is your go too!
AI answers questions in blank page

💡 Inspiration: The "TL;DR" Struggle

Let’s be real: nobody actually reads the full 50-page PDF instructions. We’ve all been there. It’s finals week, caffeine is the only thing keeping your heart beating, and you have to read a dense research paper that feels like it was written in 18th-century English. You try to read the first paragraph, zone out, and realize you’ve been staring at the same word for five minutes.

We built this because we are lazy. Well, "efficient." We wanted a tool that lets us skip the boring parts and just get the answers. We wanted a "Ctrl+F" that actually understands what we're looking for, rather than just highlighting keywords.

⚙️ How We Built It (The Stack)

We kept it simple so we could move fast. The Frontend: We used Next.js because we love React but hate configuring routers. We styled it with Tailwind CSS because writing vanilla CSS is painful. The "Brain": We hooked up an LLM (Large Language Model) API.

How it actually works

We didn't invent some new math. We just chained together some really powerful tools. You drag and drop a PDF. We break that PDF into tiny text "chunks." We turn those chunks into data the AI can read. When you ask a question, we just find the chunks that look relevant and feed them to the AI. Basically, our logic looks like this: Your PDF + Our Code= Ace the Class (Okay, that's not real math, but you get the point.)

🚧 The Pain (Challenges We Faced)

PDFs are awful: We genuinely underestimated how bad PDF formatting is. Trying to get code to read a multi-column layout without mixing up the sentences was a nightmare. We spent 50% of the hackathon just trying to strip out headers and footers. The AI lies: At first, the bot would just make things up. If we asked it about a history textbook, it would invent dates. We had to spend hours tweaking the system prompts to bully the AI into saying "I don't know" instead of lying to us.

🧠 What We Learned

Prompt Engineering is real: It sounds like a buzzword, but the difference between a good app and a bad one is literally just how you ask the AI to behave. Keep it simple: We started with complex plans for user accounts and history saving, but realized we just needed the core feature to work perfectly. Just ship it: The code is alright. The file parsing is held together by duct tape and hope. But it works, and that’s what matters.