Every brain-computer interface trades depth for safety. Moving the electrode closer to the neurons that generate a signal raises its quality enormously — but each step inward means more surgery and more tissue reaction. This simulation places electrodes at three real depths and lets firing neurons (gold sparks) prove the point directly: the signal that reaches the decoder above the head is what actually gets picked up and turned into a spike.
strength = amplitude / (1 + k·distance²) · tissueFactor(interface)
detected ⇔ strength > noiseFloor
accuracy = min( strength/(strength+noiseFloor)·100 , resolutionCap(era, interface) )
- EEG — electrodes rest on the scalp, outside skull and dura. Completely safe, but the bone and skin smear and attenuate the signal (tissueFactor ≈0.15), capping spatial resolution around 1–3 cm even in 2026.
- ECoG — a grid laid directly on the cortical surface under the skull (craniotomy required). This is how UCSF's speech-decoding work (Chang lab, 2021–2023) reached ~80% word accuracy from the speech motor cortex.
- Intracortical — needles like the Utah Array or Neuralink's flexible threads penetrate the tissue itself, sitting microns from single neurons. Best possible signal and sub-millimetre resolution, but scar tissue (glial encapsulation) degrades many arrays within 2–5 years, and implantation is a real neurosurgical procedure.
- Era slides through the real historical arc: BrainGate's 2004 96-channel Utah Array, through UCSF's 128–256 channel ECoG grids, to Neuralink's 1024+ channel flexible threads and modern high-density EEG caps.
The moving dot below the decoder approximates a cursor-control task like the original BrainGate demonstrations — its jitter shrinks as decode accuracy rises, exactly as a real subject's cursor control gets steadier with a better signal.