I Automate Myself Out of a Job (And You Should Too)

There is an old saying in engineering: “If you have to do it twice, automate it.”

I took that literally. I automated myself.

I built a system where Celery tasks wake up, check Sentry for errors, feed the stack trace to Claude, generate a fix, run the tests, and open a Pull Request.

While I sleep.

The Dream of the Laziest Engineer

We all want this. We want to be the “10x Engineer” who is actually just a “0x Engineer” with a really good shell script. But now, instead of shell scripts, we have LLMs.

The system is terrifyingly simple:

1. Sentry screams “NullPointerException!”
2. Celery catches the scream.
3. Claude looks at the code and says “Ah, you forgot to check for None, you idiot.”
4. Claude writes the fix.
5. GitHub runs the tests.

Does It Work?

Sometimes. About 60% of the time, it fixes the bug. The other 40% of the time, it hallucinates a library that doesn’t exist, imports it, and crashes the build.

But 60% is better than 0%.

The Architecture of Abdication

The core is Django + Celery + redis. Because everything in Python eventually becomes Django + Celery + redis. It is the law of thermodynamics.

If you want to see how I built this Rube Goldberg machine of self-replication, read on.

See the Code (Before It Replaces You)