# Greywater Reach — Final Prototype Evaluation

**Status:** OPERATING EVIDENCE COMPLETE; player debrief themes pending

## Recommendation

**Revise before another campaign. Do not continue this instance.**

Greywater Reach validated the central approach well enough to justify another bounded experiment, but not unchanged continuation. The strongest successes were causally legible consequences, active named characters, meaningful maintenance and internal leadership, optional contact, and reliable privacy/state controls. The largest gaps were the untested absence procedure, late arrival of direct player contact, repeated procedural neatness, and the review burden created when local outcomes and experiment-level rule or victory gates shared one docket.

## Validation findings

| Question | Finding | Evidence |
|---|---|---|
| Named characters act independently | SUPPORTED | Named actors negotiated, refused, amended, inspected, disclosed, and constrained player plans across all four communities. |
| Leadership manages rather than erases conflict | SUPPORTED WITH WATCH | Conflict and rival authority persisted, but bounded hearings and records sometimes made outcomes unusually orderly. |
| Danger and loss are causally legible | SUPPORTED | Injury, harvest loss, crossing burden, infrastructure failure, and deferred capabilities followed established warnings and choices. |
| Concision and maintenance remain competitive | SUPPORTED BY ADJUDICATION | All four communities achieved consequential outcomes through maintenance, governance, care, diplomacy, and selective refusal; player perception remains pending. |
| Absence consequences are playable | NOT TESTED LIVE | All four players submitted valid packages every turn and finished at zero misses. |
| Contact is available without forced convergence | SUPPORTED | Indirect awareness appeared earlier; direct contact occurred only after Ridgeford selected and resourced it in Turn 006. |
| Direction emerges without forced victory | SUPPORTED | Witnessed bounded authority became recognizable but was correctly declined as a victory condition. |
| Structured state reduces defects within review budget | SUPPORTED WITH REVISION | Validation, deterministic views, disclosure controls, and checksums worked; two special-review turns required six docket items and one validator counting defect needed correction. |

## Revisions for the next prototype

1. Separate local adjudication from experiment-level rule, attractor, and victory review so the ordinary decision budget stays legible.
2. Offer at least one earlier, genuinely optional player-contact opportunity with a clear refusal path; do not require contact for success.
3. Introduce conflicts that cannot all be converted into procedural safeguards within one turn, while preserving causal fairness and named human agency.
4. Exercise the absence procedure in a noncanonical fixture or consented test before relying on it in another live campaign.
5. Generate timing and decision-burden telemetry directly from status history rather than reconstructing it after closure.
6. Retain the matching-player delivery, disclosure, provenance, and deterministic-state controls substantially unchanged.

## Pending evidence rule

Raw player responses remain private. When responses arrive, archive each one byte-identically, extract only aggregate themes, and revise this evaluation if player evidence materially contradicts an operating finding. Do not quote or attribute criticism without separate permission.