Likelihood and impact are a provisional estimate, out of 100, of how likely each trend is to play out by 2035 and how far it would change national capability if it did: a weighted average of three AI readings (Claude Fable 5.1 half, two other models a quarter each), each scoring every trend separately against a written rubric. The scores will be replaced by the full model panel and then the expert elicitation. They are not used in the ranking or the network. Attention changes are read as ratios, so a doubling and a halving sit the same distance from no change. Dots with the same two scores are nudged apart. How likelihood and impact were estimated.