Every security system eventually assumes the wrong person is holding the keyboard.

History is full of moments where laptops are seized, systems are confiscated, or access changes hands faster than software can react. Most security tools implicitly trust that whoever sits behind the keyboard remains the rightful operator.

I became interested in a different question:

Can software preserve a person's intent even after they can no longer interact with it?

Wrong8007 is my exploration of that idea.

Rather than focusing on preventing compromise, it focuses on the brief window immediately after it. The system waits silently for conditions defined entirely by its owner and when those conditions are met, executes one carefully programmed action exactly once.

That seemingly simple requirement led to surprisingly interesting systems problems: where execution should live, how triggers remain isolated from policy, how ownership boundaries are enforced and how predictable behavior can be maintained even when the rest of the machine is no longer trustworthy.

The result is less a kernel module than an experiment in defensive autonomy, asking how much agency software should retain when its human operator no longer can.