I built a daily AI agent that turns a creator's new YouTube videos into organized knowledge — fully automated, running in the cloud.
What it does, every morning:
🔍 Detects new uploads on a channel (YouTube RSS)
📝 Grabs the transcript (auto-captions)
🧠 Gemini writes a German summary + hashtags
✅ Creates a ClickUp task, hashtags as searchable tags
Stack: Trigger.dev (v4, scheduled tasks) · Claude/Gemini · YouTube · ClickUp · TypeScript The fun part: real-world hardening. I hit YouTube rate-limits and a Gemini overload mid-build — so it now retries with backoff and auto-falls-back to a second model. Deployed to production, daily cron is live. 🚀
Others talk to AI. You put it to work.