Since the time of the ancients, Astronomers have been visual thinkers. It was also a hotbed of "data science" and "open science" before those were cool. This trifecta of visual, data-rich, and open creates unparalleled opportunities for new software approaches to lead to new discovery and understanding. This talk will describe how a new Python-based open-source visualization environment, "glue," enables users to bring together and visually explore several data sets without merging them. Glue is unique, and distinguished from popular related packages like Tableau, in two key ways: first, shared attributes of datasets can be "glued" together, eliminating a good deal of data munging; and second, glue allows users to make arbitrary interactive and algorithmic selections in 2D plots images and 3D volumes. Glue's ability to handle high-dimensional data makes it particularly applicable in fields like astronomy and medical imaging, and the talk will include examples from both fields. Glue is extensible via both a built-in iPython (Jupyter) terminal and via a user plug-in architecture applicable to both import (data ingest) and export (plotting, web output). The talk will conclude with speculation about upcoming challenges for glue, including: 3D selection, including augmented reality approaches; training scientists to think in 3 dimensions; live connections to related software (e.g. WorldWide Telescope; 3D Slicer); handling very large data sets; and possibilities for web-based versions. All features of glue are documented online at http://glueviz.org, which also gives links to the GitHub code repository. Glue is supported via NASA funding to Harvard, in support of the James Webb Space Telescope project.