On our search for habitable worlds, we have to account for explosive stellar flaring and coronal mass ejections (CMEs) impacting exoplanets’ surface (or cloud) habitability. In fact, these stellar outbursts are a double-edged sword: on the one hand, flares and CMEs are capable of stripping off atmospheres and extinguishing existing biology; on the other hand, flares might be the (only) means to deliver the trigger energy for prebiotic chemistry and initiate life. In this talk, I will highlight our TESS study of all stellar flares from Years 1 & 2 of the mission, driven by the stella convolutional neural network. Where manual vetting would have taken a life time, and conventional outlier detection would have missed the smallest flares, state-of-the-art machine learning approaches allow us a fast, efficient, and probabilistic characterization of flares. I will also discuss flaring as a function of stellar type, age, rotation, spot coverage, and other factors. Finally, I will link our findings to prebiotic chemistry and ozone sterilization, identifying which worlds might lie just in the right regime between too much and too little flaring. With the TESS extended mission and increased cadences (20s, 2min and 10min), stellar flare studies and new exoplanet discoveries will ultimately aid in defining criteria for exoplanet habitability.