Comment by tedd4u

"If the model predicts wrong on your filaments, that’s a data problem, not a model problem. We need more measurements."

…

"… there’s a real path to a proper perceptual colorspace model in the spirit of Mixbox or Spectral.js – predictions that are correct by construction across the full ratio space, not patches on residuals. None of this is hard. The math exists. It’s a data problem."

…

"Not in some distant future. Now. The hardware exists. The slicers exist. The filaments exist."