Full docs

The public reference for how the portal works.

These pages explain the learning model, roles, current beta scope, and safety boundaries without exposing internal prompts, credentials, or deployment details.

Learning model

AI is the obstacle, not the shortcut

Structured conversation

Students interact with an AI agent whose behavior is shaped by a challenge template. The interaction requires questions, evidence, explanations, and revision.

Separate evaluation

A judge agent evaluates progress after turns against checkpoints. A completion agent produces scores and a narrative report when the session ends.

Evidence for educators

Teachers receive a transcript, checkpoint timeline, dimension scores, hint level, completion summary, and any configured reflection response.

Student-facing boundaries

Students see the challenge brief and natural-language hints. Internal directives and checkpoint details stay hidden during active work.

Current beta scope

What is live, and what is deliberately next

Area	Current beta	Planned direction
Roles	PlatformAdmin, Teacher, InstitutionAdmin, Student, and IndependentLearner are live role-based paths.	Self-registration (built, switched off in the beta), plus independent and institution subscription tiers.
Challenge authoring	Templates of five challenge types (Socratic Constraint, The Adversary, Devil's Advocate Loop, Counterfactual Reasoning, Analogy Construction) written in plain fields, version publishing, and assignment. Independent learners write, edit and archive challenges of their own in any of those types.	The remaining challenge types, programs, community sharing of learner-written challenges, and richer customization.
Classroom operations	A teacher builds rosters and assigns an ordered chain of challenges with a narrative theme to a student or roster, inside an enforced Open At / Close At window, and can cancel an assignment to stop it at once.	Join codes, invitations, bulk imports, co-teachers, archives, and scheduling extensions.
Reports	Student and teacher session reports with transcript, checkpoints, scores, and summary.	Institution dashboards, exports, gradebook and LMS workflows.
Commercial model	Not implemented as a beta workflow.	Institution seats and independent, teacher, and student subscription tiers.

Session lifecycle

The state of a student challenge

Step 1
Start
The student opens an assigned or available challenge and begins after reading the brief.
Step 2
In progress
Every turn is saved. The judge tracks progress and hints can escalate after lack of productive progress.
Step 3
Complete or pause
The student may pause and resume. A safeguard flag pauses the session until a teacher reviews it.
Step 4
Report
Completion produces a report. Partial completion still preserves the record and relevant feedback.

Safety and integrity

Human review stays in the loop

Safeguard flags

Potential distress, self-harm, or abuse disclosures can pause a session for teacher review. A teacher can clear or escalate using the local process. Automated external alerts are not part of the beta.

Academic integrity

The product preserves reasoning evidence rather than relying on AI-writing detection. Suspicious patterns are review signals, not automatic accusations or blocks.

Role guides

Find the guidance that applies to you

Institution

Manage your institution's users and their teacher, student, or admin memberships.

Institution guide

Teacher

Build a challenge template, publish it, assign a chain to a roster, and review reports.

Teacher guide

Student

Student guide

Independent learner

Independent learner guide

Start quickly

A fast route to the guide for your role.

Get started