Current prototype: text-only signals. Next model: text + task difficulty.
CEFR.AI is transitioning from a proprietary text complexity score toward an open framework that measures both text difficulty and task difficulty. This prototype shows the text layer while the task layer and open benchmarks are in development.
Transparent criteria and public documentation to maximize trust.
Difficulty only matters when text and task are evaluated together.
Next app: calibrated tasks to collect evidence on learner ability.