Skip to content

doc.md

Latest commit

 

History

History
86 lines (56 loc) · 4.41 KB

doc.md

File metadata and controls

86 lines (56 loc) · 4.41 KB

Reverse-Engineering "Invisible Cheating App" - Natively Architecture

Everyone saw the viral stunt with "Natively," the invisible app designed to secretly ace coding interviews. I tore apart the app to see how it works, and turns out, the tech itself is genuinely interesting.

Natively Screenshot

How Natively Actually Works (Technical Breakdown)

Natively was built using Electron, a desktop app framework based on Chromium and Node.js, to create a transparent, always-on-top overlay:

  • Transparent Window (transparent: true) – This Electron BrowserWindow property ensures the background is fully transparent, showing only explicitly rendered content.
  • Always On Top (alwaysOnTop: true) – Electron's flag forces the overlay window to persistently float above all other applications, making it consistently accessible without being covered.

Here's an example code snippet:

const { BrowserWindow } = require('electron');

const win = new BrowserWindow({
  width: 800,
  height: 600,
  transparent: true,
  frame: false,
  alwaysOnTop: true,
  skipTaskbar: true,
  resizable: false,
  fullscreen: false,
  webPreferences: {
    nodeIntegration: true,
    contextIsolation: false,
  }
});
win.loadURL('file://' + __dirname + '/index.html');

Backend Communication

The overlay captures clipboard data, screenshots, or selected text and sends this information to an AI backend (e.g., OpenAI) via WebSockets or HTTP requests. This backend quickly processes and returns useful suggestions or solutions.

Screen Capture and OCR

Advanced implementations use native modules (like node-ffi, robotjs) to capture specific screen areas and run OCR (Optical Character Recognition) using libraries like Tesseract.js. This lets the overlay extract text directly from your screen.

Native Audio Intelligence (Rust)

Unlike basic screen wrappers, Natively uses a Rust-based native module (via N-API) to capture system audio with near-zero latency. This stream is processed in real-time, fed to Google's Speech-to-Text API, and converted into a "Rolling Transcript" that the AI constantly analyzes.

Conversation Intelligence (Rolling Context)

The app maintains a Rolling Context Window—a sliding history of the last few minutes of conversation. When you click "What to answer?", it doesn't just look at the last sentence; it analyzes the full context of the discussion to generate a highly relevant, strategic response.

Interface Mechanics (Quick Actions)

The frontend (React) communicates with the backend to trigger specific AI agents using 5 core buttons:

  1. What to answer?: Analyzes the Rolling Context to suggest the best next move.
  2. Shorten: Uses a refinement prompt to condense the output.
  3. Recap: Summarizes the transcript buffer into key points.
  4. Follow Up: Generates strategic questions based on the conversation flow.
  5. Answer: Accepts manual or voice input for specific queries.

Clipboard Monitoring

Electron continuously listens for clipboard changes, immediately activating AI-assisted processing whenever new text is copied.

But Here's the Catch

  • Security Annoyances: macOS and Windows can detect and restrict invisible overlays in secure or fullscreen contexts.
  • Performance Drag: OCR processes and constant clipboard monitoring can significantly increase CPU and GPU usage.

Real, Ethical Ways to Use This Tech

Roy was trolling interviews, but here's the thing—this invisible overlay tech is actually super useful:

  • Sales Copilots: Imagine having a "Wolf of Wall Street"-style playbook always ready—instantly giving your sales reps real-time context and powerful closing lines during calls or meetings.
  • Customer Support Assistant: Like having Jarvis from Iron Man whispering the perfect response into your ear—automatically suggesting accurate and relevant replies without breaking your workflow.
  • Onboarding Buddy: Give new employees a personalized overlay that pops up helpful, contextual advice exactly when they need it, helping them get productive faster and more comfortably.

License & Open Source

This project is now licensed under AGPL-3.0, ensuring that any network-based modifications remain open source.

Want to Use This for Good?

Everything's open-sourced. If this sounds like something your team could use ethically and effectively, reach out. Let's build something legit. User uploaded image 1