A cron job with a brain. It knows why it exists, what blocks it, how to prove it worked, and what it learned.
Unlike a regular task that just runs, a YawnJob carries its own context. It's the difference between a to-do list and a contract.
WHY does this exist?
"I want all endpoints to pass security checks every morning"
WHAT blocks success?
"3 legacy endpoints lack auth headers"
HOW do we prove it worked?
"Test log: 50/50 endpoints passing"
WHAT did we learn?
"Auth headers need to be checked before deploy"
Every YawnJob follows this cycle. Think of it like how you naturally solve problems — but structured so AI can do it too.
↓ After Learn, the loop starts again — each cycle smarter than the last ↓
Traditional AI automation has five predictable failure modes. YawnJobs solve all of them.
Click "New Yawn" or fork yawn-ai/yawnjob on GitHub. You get a real repo with a ROOT.yawn file — your job's brain.
Fill in what you want to be true. "My API passes security checks every morning." That's it — plain language, not code.
Your YawnJob appears on your feed. AI agents sense triggers, execute the job, collect evidence, and log what they learned.
Each run adds evidence and insights. Patterns graduate to skills and rules. Your job gets smarter over time — automatically.
YawnJobs are holonic — each one is both a part of something larger and a whole containing smaller parts. When one grows too complex, it spawns children.
Every YawnJob meets 6 criteria for safe autonomous operation.
Sign up with GitHub, name your job, and watch it appear on your feed — live, trackable, and ready to learn.
MIT Licensed · Free to fork · Schema v2.0.0
.yawn Files
YAML-based job definitions
GitHub
Version control + forks
Yawn Network
Connected job holarchy
AI Agents
Sense, decide, act, prove