Outcomes
- Repeatable transcript → summary → tagged KB entry pipeline
- Outputs land in version-controlled Obsidian vault
- Reduced time from 2+ hours manual notes to ~20 minutes review
- Structured format with key insights, quotes, and action items
Problem
Podcasts and long-form YouTube content contained valuable insights, but notes were scattered, inconsistent, and never retrieved. Listening was easy; knowledge reuse was nearly impossible.
Goal
Design a repeatable workflow: any audio/video source becomes a structured, searchable knowledge base entry with minimal manual effort.
Before / After
Before
Scattered podcast notes in random files, no tags, no retrieval, 2+ hours per episode
After
Transcript → AI summary → tagged Obsidian entry with insights, quotes, and links — ~20 min review
System
Podcast / YouTube URL
Transcript extraction
AI structured summary
Key insights + quotes
Obsidian Knowledge Base
How it works
- Capture — Save URL or audio; run through Whisper or existing transcript API.
- Structure — Claude prompt extracts: main thesis, key insights, notable quotes, action items, tags.
- Store — Output saved as markdown in git-backed Obsidian vault (
Knowledge Base). - Reuse — Entries link to trading reviews, macro research, and weekly planning.
Result
This workflow is the most marketable proof of AI implementation skill: it shows I don’t just use AI — I design systems that produce durable, structured output teams can actually use.
What I’d improve next
- Automated ingestion via RSS + webhook
- Semantic search across the knowledge base
- n8n orchestration for fully hands-off runs
- Quality scoring before auto-publish