Skip to content

Mobile-first system to capture, transcribe, and prioritize high-signal short-form content using explicit user intent.

Notifications You must be signed in to change notification settings

manthank17-learn/mindscrole-android

Repository files navigation

Mindscrole

Mindscrole is a personal knowledge intake and prioritization system designed for people who consume high-signal information through short-form content.

It helps transform scattered Instagram Reels, videos, and links into structured, searchable, and prioritized knowledge—without scraping platforms or violating user trust.


The Problem

Most valuable information today is discovered through:

  • Instagram Reels
  • Short videos
  • Social media snippets
  • Event announcements
  • Hiring posts and opportunities

But this information is:

  • Ephemeral
  • Unstructured
  • Hard to track
  • Easy to forget
  • Buried inside feeds and timelines

Mindscrole exists to capture intent at the moment of discovery and turn it into something usable.


The Core Idea

Mindscrole does not scrape platforms or automate user accounts.

Instead, it is built around one principle:

Only process content when the user explicitly chooses to.

The user decides what is worth saving.


System Architecture (High Level)

Mindscrole is composed of four core layers:


📱 Android App — Intent Capture

  • Native Android application
  • Integrated with the system Share Sheet
  • Appears when a user shares content from Instagram or other apps
  • Receives only what the user explicitly shares (usually a public URL)
  • Acts as a trusted bridge between social apps and the backend

The Android app is intentionally minimal:

  • No scraping
  • No background automation
  • No access to private data
  • Pure user-initiated action

🖥️ Backend — Ubuntu Processing Node

  • Runs on a dedicated Ubuntu machine
  • Responsible for:
    • Media handling
    • Audio extraction
    • File management
    • Transcription
    • Metadata storage

Core responsibilities:

  • Downloading shared media when applicable
  • Converting media formats
  • Running speech-to-text on short-form content
  • Persisting transcripts and references

This layer is built for stability, reproducibility, and control.


🧠 Intelligence Layer — LLM & Prioritization (Python)

On top of raw transcripts, Mindscrole applies intelligence:

  • Language understanding
  • Content categorization
  • Priority assignment
  • Signal vs noise separation

Examples:

  • Job opportunity vs general advice
  • Event with a date vs timeless content
  • Hiring announcement vs opinion
  • High urgency vs long-term reference

This layer is designed to evolve over time into a personal decision support system.


🌐 Frontend — Unified Access (React)

A web frontend provides:

  • Centralized access to everything processed by Mindscrole
  • Clean views of transcripts and extracted information
  • Categorized timelines (events, hiring, learning, ideas)
  • One place to review what the user chose to save

The frontend is the single source of truth for the user.


What Mindscrole Is Not

  • Not a scraper
  • Not an Instagram automation tool
  • Not a growth hack
  • Not a background crawler

Mindscrole only works when the user says “this matters.”


Why This Approach

  • Respects platform boundaries
  • Avoids account bans and instability
  • Aligns with real user behavior (mobile-first)
  • Scales without violating trust
  • Keeps the system legally and technically sane

Current Status

  • Android Share Sheet MVP: Working
  • Backend transcription pipeline: Stable
  • Intelligence layer: In progress
  • Frontend: Planned

Vision

Mindscrole aims to become:

  • A personal intake system for high-signal information
  • A memory extension for short-form knowledge
  • A prioritization engine for opportunities and ideas
  • A calm alternative to saving chaos

You don’t scroll less.
You lose less.


Mindscrole is built deliberately, slowly, and with intent.

About

Mobile-first system to capture, transcribe, and prioritize high-signal short-form content using explicit user intent.

Topics

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Contributors 2

  •  
  •  

Languages