Back to projects

BlinkyOwl

Zero cost, fully autonomous system with resilient self-learning built in for output to compound over time. A production-grade pipeline that generates, publishes, and self-improves educational YouTube Shorts daily with zero human intervention.

Python TypeScript Remotion Groq YouTube API

The problem

Content creation is manual, repetitive, and doesn't scale. Making even one educational video per day requires scripting, illustration, voiceover, editing, and publishing. Most creators burn out before they hit consistency. What if the entire pipeline, from idea to upload, could run itself and get better over time?

What I built

BlinkyOwl is an 8-stage autonomous pipeline that runs daily at 3 AM via launchd, producing educational YouTube Shorts for kids ages 4-7. No human touches the process. From topic selection to YouTube upload, everything is automated with built-in self-improvement loops.

Script Engine → Safety Gates → Media Generation → Video Rendering → YouTube Upload → Cross-Post → Analytics → Feedback Loop

Key capabilities

Why it matters

390 passing tests across unit, integration, regression, and end-to-end suites. The entire system runs on free-tier infrastructure: Groq for LLM calls, Gemini for supplementary generation, edge-tts for narration, Pollinations.ai for images, Surge.sh for dashboard hosting. Zero monthly cost.

BlinkyOwl demonstrates that a fully autonomous, self-improving content system can run at zero cost with compounding output quality. The self-learning loop means every day's output informs the next. Topics get sharper, thumbnails get more clickable, and engagement compounds.