Canonical · L3 Category·Shipped
AI harness
An AI harness is the orchestrated rig around a model where governance actually lives — tools, memory, permissions, review paths, escalation, outcome boundaries. In Prompted LLC's usage, harnesses are the unit governed by Ubiquity.
Plain definition
An AI harness is the surrounding system that gives a model context, tools, permissions, memory, review paths, and outcome boundaries. The model reasons; the harness governs.
Why it matters
Most discussion of "AI safety" or "AI governance" targets the model. The governable surface is the harness, not the model. Naming the harness as a distinct architectural unit is what makes earned autonomy operational.
What it is not
- A Hugging Face model release format.
- A Ray Serve deployment primitive.
- A prompt template, system prompt, or persona.
- A specific framework — harness is an architectural role, not a vendor SKU.
Where it appears in Ubiquity
Ubiquity governs harnesses. See /ai-harnesses for the full term page.
Ladder context
Demand ladder
L1 Pain→L2 Contrast→L3 Category→L4 Primitive→L5 Branded
Adjacent transitions
Related terms
Frequently asked
- What is an AI harness?
- The runtime apparatus around a model: orchestration, tools, memory, review surfaces, escalation paths. The unit governed by Ubiquity.
- How is a harness different from a model?
- The model reasons. The harness decides what the model can act on, what it must ask about, and what happens after action.
- How is a harness different from an agent framework?
- A framework is a SKU. A harness is an architectural role any framework can play with the right primitives.
- Why does governance live in the harness?
- Because the model is opaque, replaceable, and improving. The harness is where the organization actually decides what AI may do.