You're both right in part, but rabinoz is more in error, especially with his last speculations.
Google Earth uses many different sources. It is reasonable to call satellite data a primary source, as it covers the most area; including all the remote, unpopulated areas, and areas for which there is no need to have resolution of a few inches.
Cities and populated areas require higher resolution, and are imaged with aerial photography. Frequently this is already done for local government property management purposes and Google acquires that data.
Quite a lot of Google Earth imagery that falls into the "need a reasonable, but not super-high aerial resolution imagery" comes from DigitalGlobe's and SIC's QuickBird II and WorldView, WV-2, WV-3 satellites (and by now, others). As time progressed from the turn of the century, the capabilities of these instruments -- and the internationally agreed limit on commercial image resolution -- went from 0.6m panchromatic (2.4m multispectral) to 0.5m (1.5m multi-) to 0.31m (1.24m multi-).
I've worked directly with these images, and it is absolutely possible to see houses in detail (I'll leave you to quibble about "such high definition"). You can also see cars in reasonable detail, count individual people in a scene, and see sub-pixel features such as painted lines in roadways.
Landsat 8 is a poor example of the limits of high-resolution imagery available to the public.
You can most definitely purchase an image directly from a commercial satellite imagery company that shows your house in sub-meter detail.