Measure · Understand

Constellation

A constellation is one input seen through many lenses at once — the native Calyx record. Its shape across slots is the meaning, which is why the lenses are kept rather than averaged away.

A sentence is measured through eight lenses, producing eight slots. To store it, should Calyx average the eight into one vector, or keep all eight?

A sentence is measured through eight lenses, producing eight slots. To store it, should Calyx average the eight into one vector, or keep all eight?

Keep all eight. The pattern ACROSS the slots — which lenses agree, which disagree, which stay absent — IS the meaning. Averaging collapses that shape into one point and throws the information away. The constellation is the native record precisely because it preserves the shape.

A star map of one idea

Take one input. Measure it through every lens in the panel. Each lens plots a slot — a star. The full set of stars for that one input is a constellation: the native record Calyx stores.

The crucial idea: the shape is the meaning. Where lenses agree, the stars cluster; where they disagree, they spread; where a lens has nothing to say, a star is simply absent. That pattern across perspectives carries far more than any single averaged vector could.

Why keep the lenses instead of averaging them

A plain vector database would flatten those eight stars into one. Calyx keeps them, because the differences between lenses are signal, not noise — they are exactly what the later verbs read. Counting agreement needs separate slots to compare; differentiating signal needs to ask “how much does this lens add?” — a question that only exists if the lenses were never blended.

This is the whole point of the no-flatten doctrine: a constellation stays a constellation, end to end.

Practise to master it

Where this sits in the concept graph

Graph node C4 · tier F · target level Understand.

Sources