Creativity can be measured with a variety of methods including self-reports, others reports, and ability tests. While typical self-reports are best understood as weak proxies of creativity, biographical reports that assess previous creative activities seem more promising. Drawbacks of such measures -- including skewed item distributions, a lack of measurement invariance across sex, and low convergent validity with established measures of creativity (such as divergent thinking tasks) -- might be attributable to issues of item sampling. Therefore, we use a meta-heuristic algorithm ("Ant Colony Optimization") to develop a short-scale of creative activities that a) maintains coverage of the original domains, b) provides good model fit, c) has good reliability, d) demonstrates balanced item difficulty, e) is measurement invariant across sex, and f) possesses convergent validity with divergent thinking. For an 8-item short version of the original "Inventory of Creative Activities and Achievements," we identified an item set that satisfies the most pre-specified criteria using data from N = 296 adults. In a replication sample (N = 327) we replicate the findings and report good model fit, good reliability, adequate item difficulty, and measurement invariance across sex. We derive recommendations on the measurement of creative activities, and discuss implications for understanding creativity in general.