Graphs appear naturally in many applications. With ever increasing computing power available for scientific research, and popularity of online applications such as Facebook and Twitter, the size of graphs of interest increases exponentially with time.  In this talk, we discuss algorithms and challenges in visualizing large graphs, particularly recent work on scalable distance embedding.  In addition, we will look at visual representation of cluster relationship in graphs and recommender systems, in a way that is easy to understand for the general public. We then apply the same principle for visualizing Twitter text streams in real time.  The approach automatically groups similar tweets into "countries" with keyword summaries, and handles the need for visual stability across time by dynamic graph layout techniques. The resulting system, TwitterScope, provides a continuous, succinct view of evolving topics of interest.