The Capture Pattern

What we were doing

We built a YouTube to Notion workflow. It watched a playlist and saved new videos automatically.

Where it broke down

• App review: YouTube wants us reviewed before users can connect
• Token expiry: OAuth tokens expire, refresh logic breaks
• Wasted polls: Most checks find nothing new
• Playlist friction: People don't naturally add videos to playlists
• Silent failures: When something breaks, nobody knows

The idea

This works if

• We can get transcripts without YouTube API access
• Desktop and mobile both work
• It's faster than the old way
• People actually use it

How it works now

User sends us the URL. We scrape the transcript. Done.

The flow

Why this works

Transcripts are in the page HTML. Anyone can see them. We just need the URL to fetch them ourselves. No API key, no OAuth, no rate limits.

What we built

The pieces

Transcript Extraction

// No API key needed - public page scraping
const html = await fetch(`https://youtube.com/watch?v=${videoId}`);
const captionTracks = html.match(/"captionTracks":(\[.*?\])/);
const trackUrl = JSON.parse(captionTracks)[0].baseUrl;
const transcript = await fetch(trackUrl); // XML format

Before and after

Numbers

What we learned

This showed us

• • You don't need OAuth for public data
• • When someone clicks, they meant to — that's a clearer signal than automation
• • Extension plus bookmarklet plus mobile share covers everyone
• • Less code means fewer things that break

We don't know yet

• • Whether scraping stays reliable (YouTube could change their HTML)
• • Whether people prefer clicking over "set and forget" automation
• • How this holds up at scale
• • What happens when content isn't public

Trade-offs we made

• • User does more: One click per video instead of automatic
• • Fragile scraping: If YouTube changes their page, we break
• • Extension friction: Desktop users need to install something

When this makes sense

If you want to do this

The files

workway-platform/
├── apps/api/src/routes/share.ts        # API endpoint
├── apps/web/src/routes/share.tsx       # Landing page
├── apps/web/src/routes/extension.tsx   # Install instructions
├── apps/web/src/lib/hooks/useBookmarklet.ts
├── apps/web/src/components/
│   ├── CaptureToolsModal.tsx
│   └── ShareToolsSection.tsx
└── apps/extension/                     # Chrome extension
    ├── manifest.json
    ├── background.js
    ├── popup.html
    └── popup.js

The steps

1. 1. Build a /share endpoint that takes URLs
2. 2. Add scraping for your content type (YouTube, etc.)
3. 3. Make a landing page that shows progress
4. 4. Build a browser extension
5. 5. Add share_target to your PWA manifest
6. 6. Add a bookmarklet for people who skip extensions

The result

Still figuring out

• Long-term reliability: Does scraping hold up when YouTube changes things?
• User preference: Do people actually prefer clicking over automation?
• Extension installs: How many people bother to install it?