Terrene Open Academy
Terrene Open Academy publishes open-source programmes, courses, and learner tools for education in the AI era.
The Academy is organized as Lyceum, the Foundation’s education arm. Lyceum is to education what Loom is to codegen: it weaves educational content. It has three lanes, each serving a different audience.
How it works
Section titled “How it works”ATELIER (methodology authority)├── COE (CO for Education) ← how to design assessments├── COL (CO for Learners) ← how students collaborate with AI│▼LYCEUM (education arm)├── Programs ← canonical knowledge (source of truth)│ ├── ASCENT ← ML Engineering│ ├── FORGE ← CO practitioner skills│ └── Finance ← Finance domain knowledge│├── Courses ← instructor-facing (derived from programs)│ ├── Leading AI Transformation (from FORGE)│ ├── ML Curriculum (from ASCENT)│ └── ML for Decision-Makers (from ASCENT)│└── Learners ← student-facing (derived from programs) └── FNCE210 (from Finance)Programs hold the knowledge: slides, exercises, skill atoms, case libraries, datasets. Programs are the source of truth for what to teach.
Courses package program content for instructors: session plans, assessment design, rubrics. Instructors are the users.
Learner tools package program content for students: tutors, study guides, exam coaching. Students are the users.
Knowledge flows downward (programs to courses/learners). Insights flow upward (when a course discovers a knowledge gap, the fix goes into the program first).
Programmes
Section titled “Programmes”ML Engineering
ASCENTFrom Foundations to Mastery. 1,333 slides. 10 modules. 320 hours. 80 exercises with solutions. 11 Singapore-context datasets. Three teaching layers per concept: a banker and a PhD sit in the same classroom, both leave having learned something new.
Start ASCENT →CO Practitioner Skills
FORGEThe canonical practitioner library for the CO ecosystem. 67 skill atoms, 57 drill specs, 25 teaching cases, 25 drill exemplars. Sourced from 232 journal entries of real craft evidence across codegen and research workflows. Courses draw from FORGE; FORGE imposes no sequence.
Explore FORGE →Finance
Finance ProgrammeFinance domain knowledge: market theory, valuation methods, financial instruments, academic writing conventions, and case libraries. Feeds the FNCE210 learner tool (student-facing) and is powered by the COL-F methodology (24 agents, 20 knowledge areas, 13 enforcement rules).
Finance details →Methodology layer (from Atelier)
Section titled “Methodology layer (from Atelier)”Lyceum does not design its own methodology. It consumes CO domain applications from Atelier, the Foundation’s methodology authority:
- COE (CO for Education) defines how instructors design AI-aware assessments. The COE Spectrum, anti-gaming architecture, rubric design. Instructors use COE to structure the learning environment.
- COL (CO for Learners) defines how students collaborate with AI. Six-phase workflow, four approval gates, subject-layer pattern. Students use COL as their Claude workspace.
The methodology (how to collaborate with AI) is separate from the content (what to learn). A finance student uses COL-F (methodology) together with the Finance programme (content). An ML student uses ASCENT (content) directly through exercises and slides.
What all programmes share
Section titled “What all programmes share”Every Academy programme is built on CO (Cognitive Orchestration), the Foundation’s methodology for structured human-AI collaboration. That means:
- Specialized agents: Each programme has domain-specific AI agents (finance tutors, research assistants, exam coaches) that understand your field
- Session persistence: Assignment constraints, rubric criteria, and course conventions reload every session. The AI does not forget your syllabus.
- Approval gates: Students and instructors approve each phase before the next begins. AI does not proceed without sign-off.
- Accumulated knowledge: Patterns observed across sessions feed back into the system. The methodology improves over time.
- Three failure modes addressed: Amnesia (losing context), convention drift (reverting to generic behaviour), and safety blindness (missing integrity issues)
Relationship to Foundation standards
Section titled “Relationship to Foundation standards”Full content map
Section titled “Full content map”| Lane | Name | Domain | Source | Status |
|---|---|---|---|---|
| Programme | ASCENT | ML Engineering | — | Full |
| Programme | FORGE | CO + COC Skills | — | Full |
| Programme | Finance | Finance domain knowledge | — | Partial |
| Course | Leading AI Transformation | DX Training (SMU MGMT6020/6030) | FORGE | Active |
| Course | PCML | ML Curriculum | ASCENT | Full |
| Course | ML for Decision-Makers | ML for Executives | ASCENT | Minimal |
| Learner | FNCE210 | Finance companion | Finance | Active |
The Mirror Thesis underpins the methodology: assessment splits into a Trust Plane (rubric design, owned by the instructor via COE) and an Execution Plane (work production, owned by the student via COL).