The model is not the whole system. These layers answer different questions. They are not interchangeable.
Model
What intelligence is doing the thinking?
Claude · GPT · Gemini · Grok
Harness
How is that intelligence equipped to work?
Chat · Cowork · Code
Connections
What can the harness reach?
Connectors · apps · Graph · MCP
Skills
What reusable method can it perform?
Research synthesis · decision memo · stakeholder brief
Product
Which implementation am I opening?
ChatGPT · Claude Cowork · Claude Code · Codex · Copilot
Agency
How much of the work does AI carry?
Assist · Augment · Adapt
My role
Where do I remain responsible?
At the helm · in the loop · on call
The model is the intelligence. The harness determines how it can work. Connections determine what it can reach. Skills give it a reusable method.
Agency determines how much responsibility it carries. You remain responsible for the result.
Foundation beneath the stack
Timeline
What protects the work, preserves its history and lets us recover?
Sacred Timeline · captures · experiments · proposals · human review
Before AI changes important work, protect the timeline.