Wireless implantable neural recording chips enable multichannel data acquisition with high spatiotemporal resolution in situ. Recently, the use of machine learning approaches on neural data for diagnosis and prosthesis control have renewed the interest in this field, and increased even more the demand for multichannel data. However, simultaneous data acquisition from many channels is a grand challenge due to data rate and power limitations on wireless transmission for implants. As a result, recent studies have focused on on-chip classifiers (Fig. 1 top), despite the fact that only primitive classifiers can be placed on resource-constrained chips. Moreover, robustness of the chosen algorithm cannot be guaranteed pre-implantation due to the scarcity of patient-specific data; waveforms can change over time due to electrode micro migration or tissue reaction, highlighting the need for robust adaptive features.