A ranked list answers “what scores highest”. A graph answers “what sits near what”. The 3D field exists for the second question: calibration techniques cluster because calibration couples to everything; a stranded node is visibly stranded. The layout pulls high-adjacency nodes toward the centre so the eye starts where the activity is, then follows the edges outward.
Adjacency rewards what is already busy. The radar re-weights toward momentum, novelty, and feasibility — the forward-looking trio — to surface candidates that the headline ranking buries. It is the same data viewed through a different question: not “what is loud now” but “what is reachable next”.
Search and filters intersect: filtering by kind and then searching a tag narrows both the scene and the list together. Unmatched nodes dim rather than disappear, so the structure stays visible while the focus shifts.
No autoplaying camera moves, no parallax, no number that animates upward for drama. The most important affordance is that every score can be opened and read component by component.