The MEMLNauts are exploring new ways to be creative with machine learning, with impish algorithms powering freshly designed instruments . Expect fractious brass, wayward woodwind, unruly vocal processing, and indecorous string feedback. A collection of improvised performances from participants in the Musically Embodied Machine Learning project (MEML). We are exploring the concept of building new hybrid musical instruments with embedded machine learning. Embedded machine learning empowers musicians to find new performance approaches, by sharing control with often unpredictable and lively algorithms within the instrument. We will present performances with a variety of machine learning augmented acoustics instruments, in solo and group improvisations.