This is a talk I gave recently at Data Science MD. The abstract: "We all build complicated applications that do amazing things. They juggle state for thousands of objects, perform complicated algorithms across multiple threads, and stay up for weeks or months at a time. We need to understand what's happening in our applications and in the terabytes of data our applications produce. Until now the primary means of communication of internal application state to developers has been symbolic; in gigabytes of logs, alert statements, stack traces, and emails. We can do better. Jason will argue the case that we should be building visualizations to debug our code."
This was inspired by the work of Bret Victor, Chris Granger, and others.