That software exists to automatically apply changes to code doesn't have much bearing on software being able to introspect and decide what changes to apply, but maybe I'm moving the goalposts. The infrastructure exists for programs to edit their own deployment, what I'm doubting is that the software will have any idea of what can be improved without a human-in-the-loop.

Compiler optimizations are a counterexample, could be called automated self-improvement -- but it's to do the same thing with fewer resources, what I'm looking for is a reason to believe an optimization could go the other way, have more capabilities with the same resources.