Skip to content

CLAUDE.md

Latest commit

 

History

History
90 lines (66 loc) · 2.96 KB

CLAUDE.md

File metadata and controls

90 lines (66 loc) · 2.96 KB
Raw

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Repository Overview

This is a static content archive containing 303 episode transcripts from Lenny's Podcast, with an AI-generated topic index for easy discovery.

Structure

├── episodes/
│   └── {guest-name}/
│       └── transcript.md    # YAML frontmatter + transcript content
├── index/
│   ├── README.md            # Main entry point with topic links
│   └── {topic}.md           # Individual topic files (e.g., product-management.md)
└── scripts/
    └── build-index.sh       # Script to regenerate the topic index

Transcript Format

Each transcript.md contains:

  • YAML frontmatter: guest, title, youtube_url, video_id, publish_date, description, duration_seconds, duration, view_count, channel, keywords
  • Transcript content: Timestamped speaker dialogue

The publish_date field is in YYYY-MM-DD format and represents the YouTube upload date.

Index

The index/ folder contains AI-generated keyword tags for each episode:

  • Topic files (e.g., product-management.md) - Episodes grouped by topic keyword

Working with Large Transcript Files

Transcript files are large (often 25,000+ tokens). Use these strategies:

1. Use Grep for targeted searches (preferred)

# Search for specific topics across all transcripts
Grep pattern="product.market fit" path="episodes/"

# Search with context lines for better understanding
Grep pattern="early stage" path="episodes/" output_mode="content" -C=5

2. Read frontmatter first (lines 1-15)

Get metadata before deciding to read more:

Read file_path="episodes/guest-name/transcript.md" limit=15

3. Read in chunks when needed

For sequential reading, use offset/limit:

Read file_path="..." offset=1 limit=500    # First chunk
Read file_path="..." offset=500 limit=500  # Second chunk

4. Use Task tool with Explore agent

For research across multiple transcripts:

Task subagent_type="Explore" prompt="Find insights about X across transcripts"

5. Handle persisted output

When Read returns a persisted output path like: Output saved to: ~/.claude/.../tool-results/xxx.txt Read that file to access the full content.

Rebuilding the Index

./scripts/build-index.sh

This calls Claude CLI for each episode to generate keywords. The script is idempotent - it skips episodes already present in keyword files, so it can be run multiple times safely.

Adding Publication Dates

All episodes should include publish_date in ISO 8601 format (YYYY-MM-DD). To fetch the publication date for a new episode:

  1. Use the video_id from the transcript's frontmatter
  2. Call the YouTube Data API v3: 
    https://www.googleapis.com/youtube/v3/videos?part=snippet&id={video_id}&key={API_KEY}
  3. Extract snippet.publishedAt from the response
  4. Add publish_date: YYYY-MM-DD to the frontmatter after video_id