Official COG Hub

COG

Collaborative Orchestration Governance

COG is a human-AI orchestration and engineering governance system for complex AI development, aiming to advance AGI through an engineering path.

COG focuses on whether complex AI projects can become planned, executable, verifiable, reviewable, and continuously improvable project loops.

Engineering definition of AGI

AI can autonomously complete complex projects that originally required long-term collaboration by ordinary human teams.

View full definition

The Three-Axis Model of AGI Progress

01 Scale axis: determines the upper bound of local capability

02 Paradigm axis: changes capability structure and triggers leaps

03 Engineering axis: improves actual delivery capability in complex projects

AGI is the result of the three axes advancing together until system capability crosses the stable complex-task delivery threshold.

View the model

Theory

Theory and definitions

Long-form notes on COG theory, AI engineering governance, orchestration, and evaluation.

Articles

Short articles and notes

Standalone COG concept articles, practical notes, and case reflections.

Course

Systematic course

A continuously maintained AI development course under COG, covering theory, process, tools, and projects.

Latest theory

Multi-Round Audit Convergence: An External Validation Mechanism and Recursive Task Loop toward Engineering AGI

Treat MRAC as a second-order validation signal: advance on convergence, return to task decomposition on non-convergence, and build a recursive task loop toward Engineering AGI.

Read the article

What is COG

A working system growing around the engineering loop for AGI.

COG is a collection of theories, articles, tutorials, tools, development process conventions, and evidence-based data. It is not a single paper, tool, or course. It grows around the engineering loop for AI work, and its content and tools will be released as the work progresses. COG aims to advance AGI through the external engineering path.

Every step in this engineering system has practical value and can improve productivity before AGI is fully achieved. We also do not believe AGI can be realized in the short term, but that does not prevent us from advancing AGI progress from the external engineering side of models.

Releasing over time

Theory

Short articles

Systematic courses

Tools and process conventions

Evidence data

Read the three-axis model of AGI progress

Why now

AI engineering needs stronger operational governance.

AI capability is growing faster than engineering governance.

Complex engineering requires verifiable workflows.

Evaluation must become part of the production loop.

Human judgment remains central for complexity boundaries.

Learning path

Move from limits to orchestration.

Start with AI engineering limits.
Learn COG basic concepts.
Practice with evidence-based evaluation.
Move toward multi-session orchestration.

COG Eval preview

A future product area for reproducible AI engineering evaluation.

COG Eval will provide reproducible benchmark cards, runner evidence, by-case analysis, and community test submissions.

Explore COG Eval

Public Record

COG Theory Origin

Canonical definitions, version history, authorship records, and citation guidance.

Theory: Collaborative Orchestration Governance
Version: v0.1 static mockup
Record: Definitions, authorship, version history, and citation guidance.
Status: Canonical origin page placeholder