Grades over time

Every subject is re-run on a monthly cadence, and each run is kept. These are the trajectories that history builds: capability and latency per suite family, one line per agent.

Grades are comparable only within a suite version. When a suite changes, the task set changes, so the versions get a line each rather than one line sloping across the change - a v1 grade and a v2 grade are two different measurements, not two points on one trend. A suite change runs both versions for a cycle or two before the old one is retired, so the lines overlap in time and the superseded version is dashed. Plan cost is not charted here: it is a share of each agent's own subscription, so it is shown per agent on its scorecard instead of ranked across agents.