An account of how the workspace functions, including marking, mock exams, and the areas in which AI can reasonably be trusted.
The grounding, cross-checking and confidence calibration applied to every AI mark on Lumen, and how this differs from submitting work directly to Claude.