I've long maintained that CSS is one of the most well-kept (and consequently under-exploited) accessibility secrets of the Web. Thinking back to the time that CSS1 became a W3C Recommendation, those of us who cared deeply about accessibility took great care to ensure that the end-user had a lot of control with respect to how content was displayed -- yet, end-user tools that allow users to leverage this capability have been rare.
As we were experimenting with Google Reader using AxsJAX, one of the enhancements we prototyped was the addition of a simple CSS-based lens that allows the user to selectively magnify the current article. Notice that this is subtly different from using a generic screen magnifier -- in that later case, you end up magnifying the entire screen. Google Reader can be smarter; since it knows which article you are currently reading, it can selectively magnify just that article upon request. This results in much better use of screen real-estate -- something that is an even scarcer resource when you're a low-vision user.
After prototyping this via the AxsJAX framework, we decided that this feature made sense for exposing to all our users -- we have now integrated this functionality into the main Reader interface. So with this lens in hand (your pocket) you can continue to hit j and k to move through articles, and when you find the print too small to read you can press = or - to enlarge or shrink the font of the article you're reading. The C in CSS stands for Cascading -- and in this case, you the end-user get to have the final say in how you consume your content by cascading your request for a larger font on top of the presentation chosen by the content publisher.