Comet Scholar — AI Learning & Research Assistant

AI-powered notes, explanations, flashcards, and deep research—built directly inside the Comet Browser.

Comment

🧠 Inspiration

The project was inspired by the repeated cognitive load students face while studying. Every learning task—summarization, note-making, explanation, or research—requires switching between multiple tools. This creates unnecessary friction and breaks concentration.

I wanted to design a system where the learning process stays inside the browser, and every highlight becomes a meaningful output such as notes, flashcards, or simplified explanations. The Perplexity Comet Browser provided the ideal environment for this concept.

🛠️ How I Built the Project

The system was built in modular layers:

  1. Concept & Architecture

I first designed the idea, workflow, and component breakdown myself:

Highlight-based input

AI processing

Output classification (notes, explanation, flashcards, research)

Popup UI

Local storage

Structured response formats

  1. Development (Using Lovable)

Once the architecture and feature set were finalized, I used Lovable as a development assistant to accelerate implementation. The idea, logic, and behavior came from me; Lovable helped convert those ideas into working UI and backend code.

  1. Extension Layer

Manifest V3

Background scripts

Context menu actions

Multi-tab popup UI

  1. Backend (Gemini 2.5 Flash)

Created 4 AI endpoints:

Notes

ELI12 explanation

Flashcards

Smart Auto-Continue Research (SCR)

  1. UI

Reactive interface using:

React

TypeScript

Tailwind

shadcn components

📘 What I Learned

Throughout building this project, I gained experience in:

Browser extension communication pipelines

Structured prompt engineering

Designing multi-step AI reasoning chains

Creating predictable LLM outputs

Working with small UI environments (popup windows)

Managing async flows between extension → backend → UI

I also learned how to combine AI with a browser workflow in a practical, scalable way.

⚠️ Challenges I Faced

Some key challenges included:

  1. Output Stability

Ensuring the LLM always returns notes, explanations, and flashcards in a consistent structure.

  1. SCR Logic

The Smart Auto-Continue Research feature required:

Multi-hop reasoning

Topic extraction

Comparison tables

Research gaps

Diagram suggestions

Keeping this predictable was difficult.

  1. Extension Architecture

Handling the following took careful debugging:

Background → popup message passing

Triggering actions from context menu

Maintaining UI state

Local flashcard storage

  1. Time Constraints

Coordinating design, backend, frontend, and extension workflows under limited time was a major challenge.

🧮 LaTeX Example (As Required)

The project involved structuring research into logical components, similar to breaking a concept into subparts:

ResearchBundle

Summary + Insights + Comparisons + Gaps + Questions ResearchBundle=Summary+Insights+Comparisons+Gaps+Questions

This kind of decomposition helped design the SCR output format.

✅ Conclusion

Comet Scholar was built to reduce the repetitive overhead in learning. The concept, workflow, and design decisions were created entirely by me; Lovable was used to accelerate code generation and UI building.

The final system helps students learn faster, understand better, and research deeper — all without leaving the Comet Browser.

Built With

Share this project:

Updates