For decades our smart devices, from wireless sensor networks to FitBits, have generally assumed stable, reliable power from a battery or wall outlet. These devices are exploding in numbers and quickly becoming the foundational data collection platforms used to inform societal and personal scale decisions. Unfortunately, the battery dependency of these devices prevents scaling (because of bulk, expense, and maintenance) and contributes to existing global e-waste problems. This talk explores an alternative; instead of relying on energy stored in a battery, harvesting energy from the surrounding environment; however, this unstable energy supply means that these devices compute intermittently through many power failures. This new paradigm of sustainable computational things has required a rethinking of hardware, software design, and tool creation– yet it has also opened up incredible new avenues to deploy sustainable data science infrastructure, health sensing, and reimagine education at scale. I'll describe advances in making these devices more useful in the context of motivating data-science applications our lab has worked on: including late-breaking work on smart face masks, a system that enables novices to program intermittently powered devices with Python or Block-based languages, and large scale environmental monitoring.