Brief: the central prompting account

Type: types/write-brief.md

Intent

The operator's 2026-11-03–04 commission is to explain prompting through a model-defined prompt language and through English instruction execution. A KB operator should know which reference governs a failure diagnosis and what remains to be clarified or tested.

Must keep

• The adopted reference-model case must work even with deterministic execution. Ordinary English remains useful for local meaning without becoming the exact semantics of that package.
• English serves both requirement and prediction roles. Name who adjudicates the requirement while allowing the reading to remain contested; a human judge is not automatically a mechanically specified semantic reference.
• Hold the prediction question and evidence fixed when comparing inspection burdens across the two accounts. Keep requirement adequacy separate from execution conformance.

Must exclude

• Claims that English cannot have formal semantics, that prompting necessarily lacks precise semantics, or that randomness is necessary to LLM execution.
• Treating near-decomposability as the definition of opacity, or silently resolving its proposed unification with addressability. These are separate investigations.
• Assuming every prompt is governed by an adopted model reference merely because a computational account of the consumer exists.