# AWP Priority Decision — 2026-08-21

**Date:** 2026-08-21  
**Status:** Accepted owner decision. Binding until the review date below.  
**Scope:** AWP's position relative to the revenue projects, and the scope freeze that applies until AWP's one journey passes.  
**Finish line:** [`GOLIVE.md`](../../GOLIVE.md)

## 1. Decision

AWP continues as an active project alongside the revenue projects.

The reason this is not a violation of "one launch at a time" is budgetary: AWP development currently runs primarily on the owner's flat-rate ChatGPT-web pipeline. That pipeline is a fixed monthly cost with no per-token metering, so AWP does **not** compete with the metered API token budgets that the revenue projects draw on. AWP spends a resource the revenue projects cannot spend anyway.

This justification is conditional. If AWP's development shifts materially onto metered API capacity, this decision lapses immediately and AWP re-enters the normal launch queue behind the revenue projects.

## 2. What AWP is measured by

`GOLIVE.md` at the repo root, and nothing else. Percentage done is passing acceptance criteria over total acceptance criteria. Commit counts, increment gate colours, lane progress documents under `docs/progress/` and agent completion reports are not progress.

## 3. Review

**Review date: 2026-09-15.**

**Review question, verbatim:** "Has the GOLIVE journey passed end-to-end? If not, cut scope again."

The review has exactly two legal outcomes:

- The journey passes end-to-end. AWP's v1 is done; the cut list in `GOLIVE.md` is reopened in priority order.
- The journey does not pass. Scope is cut again — acceptance criteria are removed from `GOLIVE.md` and moved to the cut list until the remainder is credibly reachable. Extending the timeline without cutting scope is not an available outcome.

## 4. Standing rule — no new owner-vision extensions before the journey passes

No new owner-vision extension is specced, designed, mocked up or estimated before AWP's one journey passes end-to-end. This explicitly freezes further work on `docs/VISION-EXTENSIONS-2026-08-20.md` and the extension designs derived from it, including the Advisor personas and Cmd+K search work in `AWP-ADVISORY-PERSONAS-AND-SEARCH-DESIGN.md` and `AWP-ADVISOR-SEARCH-DEPENDENCY-DECISIONS.md`.

Existing specs stay where they are. They are not deleted, not amended, not extended, and not used to justify implementation work. New ideas are captured as one-line entries on the `GOLIVE.md` cut list and nothing more.

The rule applies to design and specification effort, not only to code. Writing a design document for a deferred capability is the exact failure mode this rule exists to prevent: it feels like progress and moves no acceptance criterion.