Operator Economy of Effort — Why this is canon

Lifted from AGENTS.md § OPERATOR ECONOMY OF EFFORT § Why this is canon, not preference under GHI #327. The operative claims and anti-patterns remain canonical in AGENTS.md; this page holds the justification narrative.

Why the design-dialogue mode is canon

The draft → review → decide → attest interaction shape produces witnessed, attestable, replayable governance work without consuming the operator's bandwidth as the bottleneck.

Every other mode either shifts the typing burden onto the operator or produces output the operator has to re-author into the canonical shape after the fact:

• Operator drafts, agent reviews. The substantive labor lands on the human; the agent's review pass is the cheap part. The economy is inverted.
• Bundled clarifying-question intake. The operator answers n questions in one round; the agent's next draft incorporates them but the operator's typing budget has already been spent on raw prose.
• Open-ended brainstorming without decision-shaping. The agent generates options without forcing-functions or bounded answer space; the operator must impose the structure after the fact.

Both inverted modes are forms of vibing through the interaction layer — the agent appears active while pushing the substantive labor onto the human or producing artifacts that need to be redone.

Relationship to the rest of the contract

The Operator Economy claims are downstream of the anti-vibing mantra and upstream of the Attestation discipline:

• Upstream (anti-vibing): "every option is framed by which choice leaves the smallest surface for vibing to leak through" — the design- dialogue mode is the interaction-layer instantiation of that frame
• Downstream (attestation): the operator's verbatim phrasing preservation rule (operative claim 3) is the same rule as the attestation-text pattern (<user verbatim words> — <session-grounded enrichment>). One originates the operator's words; the other preserves them through to the receipt.

When the design-dialogue mode does not apply

The mode binds for design, decision, doctrine authoring, ADR ceremony, and OBPI walkthrough. It does not bind for routine code edits where the operator's intent is already encoded in the brief, the failing test, or the literal user message ("fix X by doing Y"). In those cases the agent executes against the existing intent surface; the design-dialogue mode would be ceremony overhead.

Origin

GHI #327 — instructions-files-diet pass that lifted this rationale out of AGENTS.md to economize per-turn context weight while preserving the operative claims at canonical home.