The problem Cognivo solves

Got it — here's an updated version incorporating your newer features (Time Reality Check, Conflict Detection, Node Detail Editor, Persistent Canvas) and adjusting the items that may not match what's actually built:

The Problem
Students and knowledge workers constantly struggle with information overload — they consume content from YouTube, attend lectures, and juggle dozens of tasks, but have no unified way to capture, connect, and prioritize it all.
Cognivo (Visual Second Brain) solves this by giving you an AI-powered visual canvas where your knowledge, goals, and tasks live together as connected nodes.

1. Brain Dump → Eisenhower Matrix

Paste or speak your chaotic thoughts, and AI (Gemini 1.5 Flash) instantly sorts them into the Eisenhower Matrix (Urgent/Important quadrants), helping you focus on what truly matters.
2. YouTube → Actionable Roadmap

Paste any YouTube URL and AI extracts the transcript and distills it into a clean 4–8 step roadmap, added directly to your canvas as connected task nodes — so you stop re-watching videos and start acting.
3. Voice Input

Speak your thoughts naturally using voice-to-text (Web Speech API) — Cognivo transcribes and feeds your ideas directly into the AI sorting pipeline, no typing required.
4. Visual Canvas with Goal, Milestone, and Task Nodes

All your goals, milestones, and tasks become draggable, connectable nodes on an infinite React Flow canvas — giving your second brain a visual, spatial structure instead of a flat, forgettable list.
5. Conflict & Time Reality Detection

Cognivo scans your canvas to flag overlapping deadlines and illogical dependencies (e.g., a prerequisite task scheduled to finish after the milestone it feeds into), and warns you when your total planned workload exceeds the time you actually have before a deadline.
6. Node Detail Editor

Click any node to edit its label, description, deadline, and priority, mark it complete, or delete it — keeping every part of your plan up to date.
7. Search & Focus

Quickly search across canvas nodes by label or description to find and jump to any task, goal, or roadmap step.
8. Persistent Canvas

All nodes and edges auto-save to localStorage, so your second brain is exactly where you left it every time you return.

Challenges we ran into

1. YouTube Transcript Extraction: Many YouTube videos don't have captions enabled, causing the youtube-transcript library to fail silently. We added proper error handling and user-friendly fallback messages to handle this gracefully.

2. Gemini Response Parsing: The Gemini API sometimes wraps JSON in markdown code blocks (json ... ) even when instructed not to. We had to add a cleanup layer to strip markdown formatting before parsing, to prevent JSON.parse failures.

3. Real-time Voice Input: Integrating the Web Speech API for voice-to-text brain dumps was tricky — interim transcripts, multi-chunk appending, and browser compatibility (Chrome/Edge only) required a custom useVoiceInput hook with careful state management.

4. Local vs AI Classification Fallback: Ensuring a seamless fallback from Gemini AI classification to local keyword-based Eisenhower sorting (when API is unavailable) without any UX disruption required careful error boundary design.