Context engineering vs. harness engineering
Context makes the world legible; the harness makes action dependable.
Context engineering asks what information should reach the model, in what form, and at what time.
Harness engineering asks a wider question: what system turns model capability into dependable work?
Context is part of the answer. So are tools, execution environments, permissions, state, scheduling, evaluators, recovery, and feedback.
The distinction becomes clearest in long-running work. An agent may perfectly understand that it should inspect a rollout every thirty minutes. Understanding does not create a durable timer. It may know that production mutation is forbidden. Knowledge does not create an access boundary. It may write an excellent report. Prose does not make the underlying evidence reproducible.
Context helps the agent reason. The harness connects reasoning to reality—and makes the consequences governable.