The difference is that the minimal situation needed for magenta/purple is objectively different from the minimal situation needed for yellow. (Assuming reality exists, etc.)

While you could try to engineer a more efficiently yellowness-detector, engineering a better magentaness-detector would be a qualitatively different kind of biological machine.

Nature has already engineered a dedicated yellow-detector. And it qualitatively differs from what us normies think of as "yellow". Pseudo-yellow is bimodal in the same way that purple is bimodal, while true-yellow is monomodal.

Suppose humans experienced 12 primary colors. Is a combination of the 2nd, 5th, 6th, and 12th "on the spectrum" any more than an exclusive combination of the 2nd and 3rd?

You're looking in the wrong direction, you can always add a trivial trace of some additional wavelength and have no measurable effect on what people experience. That remains true no matter how many finite sensors you imagine they have. That part is simply not interesting.

What's interesting is differences in the MINIMAL set of physical things that need to be detected to get the experience.

The objective truth is that if I hand you a machine that generates just single wavelengths of photons, you can cause "Normies" to say "yellow" but you can't do the same for "magenta".

That asymmetry remains even when you talk about adding more sensors. You can frequency-shift an R/G/B cone to get a new kind that "peak-triggers when normal people say yellow" but you cannot do the same to create one which "peak-triggers when normal people say magenta". You would need to create a fundamentally different kind of sensor.

You're assuming that yellow is permanently bound to a single particular wavelength. What if it's possible to rebind yellow to a different wavelength? What if it's possible to rebind purple to a single wavelength (rather than dual wavelength)? Color doesn't happen in the eyes, it happens in the brain.

Mirimir wouldn't feel surprised about the absence of purple if they understood color in the context of phenomenology (rather than engineering), which is the spirit in which the post was made.

I understand the difference you are pointing to. It's not a novel or interesting position, it's the default position. If you insist on binding colors to wavelengths, you probably didn't get anything out of the blogpost.

And what does this have to do with colors being spectral? If you look at the cie chromaticity diagram you can see that yellow and cyan have wavelength corresponding to them while purple does not.

You probably think I'm being pedantic. But the point of the article is to conceptually distinguish color from wavelength. If we're going to distiniguish them, we might as go all the way. There's no reason to bind yellow (and cyan) to a single wavelength, while magenta gets special status. All three are secondary colors and all three are composites.

This is important because privileging magenta does not generalize to higher-dimensional color-spaces, and therefore impedes to our understanding of phenomenology.

The RGB model that is currently used (sRGB) is just an emulation of CRT-displays, it is not how human vision works. There is not actually much conceptual difference between red and yellow.

The model is irrelevant to my point. E.g. Lab and Cieluv may become obsolete if it turns out that we can remap "the bindings between color and wavelength" in our brains as easily as "the bindings between color and wavelength" in our vision libraries. The fact that purple doesn't associate to a wavelength should be a clue that binding the two at all is a mistake, not a clue that purple is exceptional.

Yellow, cyan, and magenta don't correspond to particular wavelengths. They each correspond to a combination of two particular, adjacent wavelengths.

E.g. a "yellow" wavelength is only as bright as its decomposition into red and green. The fact that red & blue aren't adjacent is a red herring.

Yellow is about 575nm, while cyan is 490nm if this diagram is correct: https://en.wikipedia.org/wiki/File:PlanckianLocus.png