This article describes an extended collaboration between a development team and an evaluation team working with GenScope, an open-ended exploratory software tool. In some respects, this was a routine evaluation, documenting substantial gains (of roughly 1 SD) in genetics reasoning ability in all but 1 of 17 classes, despite challenges presented by school computer-lab settings. Relative to matched comparison classes, larger gains were found in technical biology and general science courses but not in college prep or honors biology courses. In other respects, our effort illustrates the value of new views of assessment, technology, and research. The alignment of a sophisticated research assessment and simple classroom assessments shed light on initial failures, spurring revision. By refining the GenScope activities and extending the classroom assessments, we supported worthwhile whole-class discourse around the shared understanding of the software. A follow-up study in a laptop-equipped classroom yielded the absolute and relative gains (3.1 SD and 1.6 SD) that proponents of such innovations have long promised. In retrospect, the strengths and weakness of the study illustrate the value of newer "design-based" approaches to educational research. (Contains 11 figures, 2 tables and 14 notes.)