Will Crichton

A hypothesis: for live programming to scale beyond toy programs, generic low-level visualizations aren't enough. Instead, we need Live Literate Programming: better tools for creating *program-specific* views, which communicate our higher-level understanding.

Bearish on program visualization right now. People have been trying to vis code for half a century, and nothing has worked. All we have are Doxygen inheritance diagrams. My guess: without a more principled understanding of program comprehension, we can't build effective vis.

First, you understand the tasks. Then you use cognitive principles to match the needs of the tasks to the best representation. For example, data visualization has done this well. We know why pie charts are bad, when to use a line vs. scatter plot, so on.