Why I built it
My work at Longhorn Neurotech made me want to learn more about how neural recordings are analyzed at the level of individual neurons. The lab wanted a 3D view because it lets them compare two clusters across three principal components at once, instead of piecing that together from several 2D plots.
The hard part
The first obstacle was getting phy to run on my Mac. Its older graphics code wouldn't start on Apple Silicon, so I forced an OpenGL 2.1 compatibility profile on macOS and fixed a crash caused by a newer version of joblib. The other difficult part was the lasso split: it only checked the spikes drawn on screen, which are a subsample, so splits were missing spikes. I changed it to reload every spike before splitting.