Note On Steroids

Don't just record the lecture—understand it in real-time. Notes On Steroids is an AI-native workspace that bridges the gap between physical classrooms and virtual meetings.

Comment

Note On Steroids (NOS)

Inspiration

Over the past year, a classmate introduced me to NotebookLM, which completely reshaped how I studied. Being able to upload course materials and instantly extract structured summaries, key concepts, and explanations helped me understand subjects more deeply and efficiently. However, there was an unintended side effect: I gradually stopped taking traditional in-class notes.

Over time, I realized that live note-taking is still essential—not just for recall, but for engagement. Writing and organizing notes during lectures helps reinforce learning in ways post-hoc summaries alone cannot. I began searching for an application that could bridge this gap: something that would listen with me during lectures, capture the important parts in real time, and augment my understanding without distracting me.

Most live transcription tools I tested were either too basic (raw transcripts with no insight) or too heavy-handed (over-summarizing without context). That gap inspired Note On Steroids (NOS)—an AI-powered, real-time study co-pilot that blends live transcription with deep semantic understanding using Gemini 3.

What It Does

NOS acts as a real-time learning companion during lectures, meetings, or study sessions:

  • Captures live audio using the browser’s built-in Speech-to-Text engine.
  • Streams and batches transcripts in near real time.
  • Sends structured transcript segments to Gemini 3 for analysis.
  • Receives:
    • Concise summaries
    • Key points and definitions
    • Concept explanations
    • Structured, readable study notes

This allows users to stay focused on listening and thinking while NOS continuously builds high-quality notes in the background.

Gemini 3 Integration (Core of the Project)

Gemini 3 is central to NOS and powers all intelligence in the application.

  • Context-Aware Summarization
    Gemini 3 processes rolling transcript windows, maintaining context across time instead of treating each chunk independently. This enables coherent summaries that reflect the flow of a lecture rather than isolated statements.

  • Semantic Structuring
    Raw transcripts are transformed into organized sections (topics, subtopics, bullet points), closely mirroring how a human student would structure notes.

  • Insight Extraction
    Gemini identifies emphasis cues, repeated concepts, definitions, and examples, surfacing what actually matters for studying.

  • Low-Latency Iteration
    The Gemini API is called incrementally, allowing NOS to update notes continuously without overwhelming the user or the model.

Without Gemini 3’s long-context understanding and reasoning capabilities, NOS would be little more than a transcription app. Gemini turns speech into learning material.

How We Built It

NOS is a full-stack TypeScript application designed with clarity, security, and scalability in mind:

  • Frontend & Backend:
    A Single-Page Application served by a Node.js backend, delivering static HTML and APIs.

  • Input Validation & Safety:
    End-to-end schema validation using Zod, ensuring all user inputs and Gemini responses are sanitized and predictable.

  • Database:
    Cloudflare D1 for fast, lightweight storage of notes, transcripts, and generated flashcards.

  • API Design:
    Well-documented endpoints using Swagger, making future expansion and collaboration straightforward.

Challenges We Ran Into

The biggest challenge was platform constraints. While the browser-based Speech-to-Text works well for English demos, it lacks the accuracy, multilingual support, and offline capabilities of native mobile speech engines. Building native apps for iOS and Android introduces additional complexity but is necessary for the next phase.

Another challenge was balancing latency vs. insight quality—finding the sweet spot where Gemini receives enough context to be useful without delaying feedback.

Accomplishments We’re Proud Of

  • Successfully combining live transcription with real-time AI reasoning.
  • Designing a clean, intuitive UI that doesn’t distract during lectures.
  • Implementing robust validation and error handling across the stack.
  • Building a genuinely useful tool that solves a real learning problem.

What We Learned

  • Practical best practices for using the Gemini 3 API efficiently.
  • How to manage long-running, context-heavy AI interactions.
  • Advanced input validation and secure API design.
  • Translating product ideas into polished, maintainable software.

What’s Next for Note On Steroids

The next major milestone is mobile-first support:

  • React Native apps for iOS and Android.
  • Integration with Apple Speech and Google Speech-to-Text for higher accuracy and multilingual support.
  • New Gemini-powered features:
    • Collaborative study sessions
    • Question generation and self-quizzing
    • Post-lecture review and exam prep modes

NOS started as a personal productivity experiment and has evolved into a powerful AI-driven study companion. With Gemini 3 at its core, it has the potential to fundamentally change how students learn in real time.

Built With

Share this project:

Updates