That said, automatically annotated screenshots would be useful, or even better: if someone invents a screengrab tool that would semantically tag all parts of the image based on the data available so its not just a plain image. I.e. screengrab a GUI window and all the text displayed in that window could be added as metadata like titlebar/button/label text and then it becomes searchable. I guess it would need an API that the developer of the app would have to explicitly implement though for most useful results like debug state information...