We are hitting the thermal noise floor with neural electrodes—and smaller electrodes will not help.
Electrode SNR has not improved in 20 years despite massive materials engineering. The limit is not the electrode—it is the Johnson-Nyquist noise of the tissue itself. At room temperature, 10 kHz bandwidth, and 500 kΩ electrode impedance, you cannot do better than ~5 µVrms. Most neural signals are 50-200 µV. We are already within an order of magnitude of the physical limit.