Which of your colours stop being different
A simulated palette is hard to judge. The list of pairs that collide is the useful output, and that is what this leads with.
Which of your colours stop being different
A simulated palette is hard to judge if you have typical colour vision. What is useful is the list of pairs that collide — and that is what this leads with.
Pairs that collide
Side by side
The model
Dichromacy uses the Viénot, Brettel and Mollon (1999) LMS projection, computed in linear-light sRGB: the colour is converted to cone responses, projected onto the plane the missing cone type collapses to, and converted back.
Reduced severity is a linear blend toward the dichromatic result. That is a documented approximation, not a separate published model, and it is labelled as such rather than presented as anomalous trichromacy.
Achromatopsia preserves relative luminance, so it doubles as a greyscale check: colours that merge here are being distinguished by hue alone.
A collision is a pair whose perceptual distance falls below a threshold under simulation while being above it before. Distance is measured in OKLab.
Do not rely on hue alone
Red for error and green for success is the canonical failure. For roughly one man in twelve those two are close to the same colour, and the meaning disappears entirely.
The fix is not a different red. It is a second signal: an icon, a label, a position, a difference in lightness. WCAG 1.4.1 says exactly this — colour must not be the only means of conveying information.
Greyscale is the quick test
Set the type to achromatopsia. Anything that merges there is being told apart by hue alone, and that is the set of decisions to look at first.