The Learner Profile

Why depth

Without L2, the Mentor generates generic experiences. With L2, it generates this experience for this human.

The architecture chose depth deliberately. A profile rich enough to teach a child well is also rich enough to surveil her for a lifetime. Shallow profiles produce shallow tutoring; the cost of going deep is that L4 becomes load-bearing. L2 and L4 cannot be separated — the depth of one forces the existence of the other.

The compression pyramid

The profile is organised as a compression pyramid. At the base sits raw data: clicks, hesitations, typing speed, answer patterns, captured continuously. One level up, real-time interpretation infers fatigue, frustration and attention in seconds. Next, daily synthesis compresses thousands of micro-cognitive signals into a sharpened understanding. At the top sits the long-term profile — learning type, cognitive strengths, persistent blockades, deep interests — synthesised over months and years.

What rises through the pyramid loses none of its specificity but gains structure. The Mentor at any moment can read the top of the pyramid for orientation, and reach down to the base when a particular moment needs context.

Episodic memory and lifelong continuity

Separate from the pyramid, the system maintains an episodic memory — curated key moments the Mentor recognises as significant. "November 12th, cracked vector calculus." "The crisis in March and how it was overcome." A good mentor remembers.

The profile has no expiration date. It accompanies the person from childhood literacy through adult reskilling through retirement language learning. The same profile that learned to read at six helps prepare for a career change at forty-three and follows a sabbatical year of pottery at sixty-eight. There is no "graduation" that resets the file.

The personal competency graph

Within L2 lives a personal projection onto the Federated Knowledge Graph (L5). In the old system, a student has one position: grade seven, GPA 2.3. In this system, the learner has thousands of positions in a multi-dimensional space — algebra but not fractions, Python but not the logic underneath, physics concepts but not formula rearrangement. Granular, not linear.

From this projection, the system derives a living effort prognosis. It knows the learner, the graph and the aggregated experience of millions of trajectories. "Route A takes an estimated 24 months. Route B uses your strength in visual-spatial thinking and takes 14." The estimate updates with every step and is honest about its own uncertainty.

170 million students. No one decided.

Reference

Architecture paper, Section 5, L2. DOI: 10.5281/zenodo.18759134. CC BY 4.0.