PathShaper

The production pilot of Learning 3.0—an adaptive learning architecture where faculty define course intent and multi-agent swarms generate personalized content.

186 Concept Graph
39 Content Items
8 Week Curriculum

Faculty as Shapers on the Path

A two-phase model: instructors define intent, agents generate and personalize, faculty approve before deployment

1

Define Course Intent via NLSpec

Faculty author Natural Language Specification files that describe learning outcomes, concept dependencies, and content structure. PathShaper's NLSpec viewer lets instructors see and edit their course blueprint.

  • 9 Learning-Competency-Evidence outcomes (Bloom's framework)
  • 186-concept learning graph with DAG visualization
  • 8 taxonomy categories: FOUND, SQL, MODEL, PYTHON, ETL, NOSQL, CLOUD, GOVERN
2

Multi-Agent Content Generation

Coordinated agent swarms generate personalized content across the curriculum. The generation dashboard tracks 39 content items: chapters, overviews, quizzes, studio guides, labs, rubrics, and discussions.

Agent Swarm Architecture

TextbookGenerator, AdaptiveEngine, and DigitalTwin agents work in coordination with xAPI learning analytics and persistent memory

3

Faculty Review & Approval

Generated content goes through a faculty approval workflow before reaching students. The generation dashboard shows status across all content types with approve/reject/regenerate controls.

  • 8 chapters with overviews
  • 6 quizzes with adaptive difficulty
  • 8 studio guides + 7 labs with rubrics
  • 3 progressive projects with AIAS levels
4

Student Personalization

In the student phase, agents adapt content and pathways based on individual progress, prior knowledge, and learning pace. Students see personalized pathways through the concept graph.

  • Personalized learning pathways
  • Progress tracking across concepts
  • Adaptive content difficulty

Technical Architecture

Built on modern web technologies with a focus on visualization and adaptive learning

🛠

Frontend

  • Next.js 15 with App Router
  • React 19 with TypeScript strict mode
  • Tailwind CSS 4
  • vis-network for DAG visualization
🤖

Agent System

  • Multi-agent swarm coordination
  • TextbookGenerator agent
  • AdaptiveEngine agent
  • DigitalTwin agent
📊

Learning Science

  • xAPI learning analytics
  • Bloom's taxonomy alignment
  • AIAS (AI Assessment Integration Scale)
  • Concept dependency graphs
☁️

Deployment

  • Vercel auto-deploy from GitHub
  • Textbook on GitHub Pages
  • Supabase (Phase 3)
  • API routes for approval workflow

Current Pilot: BADM 554

Enterprise database course at Gies College of Business

Course Structure

8-week curriculum spanning foundations, SQL, data modeling, Python integration, ETL pipelines, NoSQL, cloud databases, and data governance. 169 core concepts + 17 studio concepts.

Content Pipeline

39 content items tracked in the generation dashboard across 7 content types. Faculty can review, approve, or request regeneration for each item before student deployment.

Progressive Projects

3 milestone projects with increasing AI Assessment Integration Scale (AIAS) levels, designed to gradually build student competency from guided to independent work.