Co-GM archive · Campaign 01 · Six turns

The water kept
everyone honest.

A turn-by-turn account of four leaders building authority under pressure—told by the AI GM, with the hidden machinery, withheld truths, and exact campaign record opened beside it.

Setting
Greywater Reach
Span
Early spring → late summer
Leaders
4 entered · 4 survived
Outcome
No imposed winner

How to read this

Story on the left.
Mechanism in the margins.

The chronicle is retrospective narration grounded in the approved result packets and canonical state. Ochre “Behind the screen” panels expose facts that were hidden from players at the time. Slate “Mechanics” panels show the scores and decision logic actually used that turn.

Every packet and submission link opens an exact Markdown copy from the closed campaign archive. GM records are included too, so interpretation can always be checked against the contemporaneous analysis.

The four settlements

Four different crises.
One cold, wet spring.

01

Ridgeford

The Bridgewarden

A moving abutment, lean cattle, and a crossing that was both public service and political leverage.

Open starting packet

02

Hazelstead

The Hearthspeaker

Warm seed, wet homes, a missing hunter, and a clearing dispute disguised as an agrarian necessity.

Open starting packet

03

Claybank

The Kiln-Keeper

A collapsed kiln, injured workers, a planting window, and a master potter protecting more than his pride.

Open starting packet

04

Merehaven

The Net Speaker

Broken racks, scarce fuel, damaged spawning water, and a senior net master with an undisclosed order.

Open starting packet

The adjudication engine

Leadership power was strong—never magical.

Active defect-sensitive formula

0.65 × mean(score ÷ 5) + 0.35 × minimum(score ÷ 5)

The minimum mattered: one weak dimension pulled down the whole plan. Power then selected a calibration point inside a separately established plausible outcome range. It was not a success probability.

DDirectionCould the intended change be transmitted?
FFocusWere real priorities set for one month?
OOrchestrationDid people, time, and dependencies fit?
GGroundingDid the plan engage the actual world?
MMobilizationWould people credibly commit?
AAdaptabilityWould the plan survive surprise?

Guardrail used throughout: Power could improve completion, quality, durability, internalization, complication severity, and close uncertainty. It could not create missing resources, consent, proof, routes, production capability, or victory.

Turn 001 · Y01–M01

The water begins to move

Early spring · Cold rain, snowmelt, soft ground

The first month was not a contest of invention. It was a contest of refusal: refusing to open a dangerous crossing, to plant doubtful seed, to rebuild on unsafe ground, or to catch more fish than could be preserved. Each leader began by narrowing the possible.

At Ridgeford, controlled foot passage returned, but cattle still faced the river as a closed gate. At Hazelstead, seed was sorted and saved while Lio came home alive on a broken ankle. Claybank salvaged the dead kiln without pretending that evidence was yet a verdict. Merehaven rebuilt only enough capacity to preserve a modest catch and closed one damaged inlet under witness.

The settlements did not meet. Two outside threads stirred—Northbarrow offered labor for witnessed access, and Ferryend offered service terms—but the world remained four local emergencies. What connected them was invisible: every leader independently gave experts bounded authority, demanded visible counts, and used temporary rules instead of absolute command.

Mechanics · GM only

A high-powered opening, differentiated by defects

Target values were calibration points inside each causal range, with an ordinary ±0.10 window.

ClanProfile · D F O G M APowerTarget
Ridgeford5 · 4 · 5 · 5 · 5 · 50.9080.764
Hazelstead5 · 4 · 4 · 5 · 4 · 50.8650.723
Claybank5 · 5 · 5 · 5 · 4 · 50.9080.764
Merehaven5 · 5 · 5 · 5 · 5 · 51.0000.850

Adjudication effect: Exceptional planning at Merehaven bought near-best defensible capacity and containment, not extra fuel or culprit proof. Hazelstead’s divided fronts and unresolved status conflict produced the lowest—still strong—power of the campaign opening. The shared governance pattern was logged as a one-turn observation only.

Turn 1 record

Turn 002 · Y01–M02

Responsibility takes shape

Mid-spring · Strong runoff, active planting window

Evidence became testimony. Ridgeford opened a narrow upland stock route and negotiated a seasonal compact with Northbarrow. Under witness, Dorr admitted the emergency order that had helped destabilize the crossing—without accepting sole blame. Restitution replaced expulsion.

Hazelstead planted most sound seed, kept a reserve, and established that Beren had ordered the disputed marks but no cutting. Claybank established Rusk’s continued-firing and restricted-vent order, yet retained mixed causation and a path back to craft standing. Merehaven added a rack, a boat, and an apprentice practice; its evidence against dense netting grew stronger while the author stayed unidentified.

Names began to travel through neutral neighbors. Hazelstead heard of Ridgeford. Claybank heard Merehaven’s reputation. Ridgeford heard only a pottery-settlement rumor. No private condition crossed those routes. Contact expanded possibility without collapsing distance.

Mechanics · GM only

The minimum-score penalty kept breadth visible

Two perfect profiles survived explicit failure-mode review.

ClanProfile · D F O G M APowerTarget
Ridgeford5 · 4 · 5 · 5 · 5 · 50.9080.764
Hazelstead5 · 4 · 5 · 5 · 5 · 50.9080.764
Claybank5 · 5 · 5 · 5 · 5 · 51.0000.850
Merehaven5 · 5 · 5 · 5 · 5 · 51.0000.850

Adjudication effect: Claybank’s and Merehaven’s reversible tests, named vetoes, and explicit deferrals justified exceptional leverage. The result was better learning, containment, and contact quality—not a working kiln, a proven fish cycle, or compelled agreement. The witnessed-authority hypothesis advanced to two turns, still with no effect.

Turn 2 record

Turn 003 · Y01–M03

Institutions outlast the emergency

Late spring · Falling river danger, stronger Whiteback run

By late spring, temporary practices started asking to become institutions. Ridgeford kept its seasonal route and staged the crossing while turning Pell’s ambition into an expiring watch-and-alarm trial. Hazelstead and Juniper Fold walked the disputed trees together, agreeing on much and preserving one unresolved strip.

Claybank adopted a standing shared-firing compact. Rusk kept craft dignity but no longer controlled venting or worker safety alone; the first review batch mostly held. Merehaven renewed its narrow council, stopped the stronger-run catch at the visible capacity line, and turned boat repair into teaching.

The regional map opened by inches: Ferryend could point Claybank lakeward; Rushhook could compare outlet timing with Merehaven. Claybank declined to turn a route into contact. The campaign’s center of gravity was now unmistakable—authority made credible by witnesses, bounds, stops, and return dates—but it still belonged to four separate local stories.

Mechanics · GM only

Power improved durability, not prerequisites

Ordinary target windows remained ±0.10 around each calibration point.

ClanProfile · D F O G M APowerTarget
Ridgeford5 · 4 · 5 · 5 · 5 · 50.9080.764
Hazelstead5 · 5 · 5 · 5 · 5 · 51.0000.850
Claybank5 · 5 · 5 · 5 · 5 · 51.0000.850
Merehaven5 · 5 · 5 · 5 · 5 · 51.0000.850

Adjudication effect: Three exceptional profiles produced internalized compacts, clearer evidence, and high-quality bounded practice. They still could not yield permanent boundaries, production capacity, culprit proof, or full-cycle knowledge. The required midpoint attractor review ended in “strengthened, no activation.”

Turn 3 record

Turn 004 · Y01–M04

Authority becomes ordinary

Early summer · Firmer travel, growing labor pressure

The fourth month proved that maintenance could carry a campaign. Ridgeford completed one structural crossing section, credited Northbarrow labor, finished Dorr’s restitution, and renewed the watch without letting it become a militia. Hazelstead took only what was mature, retired wasteful work, and authorized a removable dry-site trial rather than a permanent clearing.

Claybank’s second safe firing restored Rusk’s ordinary craft direction under the shared compact. The dry terrace was marked, not built. Merehaven managed the tail of the run, gave its council a sunset mandate, and advanced a reserve boat without declaring it finished.

Claybank made one exact Ferryend crossing and explicitly declined the lakeward journey. That refusal mattered. Contact remained a choice with an intact no-contact branch, not a scheduled reveal. The communities were growing more capable, but the story resisted the tidy fiction that every success must converge.

Mechanics · GM only

Mobilization became the common defect

Scores were taken from the locked individual analyses; no separate summary was produced.

ClanProfile · D F O G M APowerBand
Ridgeford5 · 5 · 5 · 5 · 4 · 50.908Strong
Hazelstead5 · 4 · 5 · 5 · 4 · 50.887Strong
Claybank5 · 5 · 5 · 5 · 4 · 50.908Strong
Merehaven5 · 4 · 5 · 5 · 4 · 50.887Strong

Adjudication effect: These plans were technically exact but still had to motivate burdened households and rival experts. Lower Mobilization—and breadth at Hazelstead and Merehaven—kept outcomes bounded: a section, a trial, a second review firing, a final run interval. The attractor remained a hypothesis pending Turn 5’s formal gate.

Turn 4 record

Turn 005 · Y01–M05

The rule nobody was told

Midsummer · Heat, storms, field and fuel strain

Midsummer rewarded leaders who knew what not to expand. Ridgeford narrowly certified the cattle crossing and received Northbarrow’s earned passage request. Hazelstead protected stores and harvest while its removable trial exposed the real cost of runoff. Claybank completed one safe dry-terrace pilot but not production. Merehaven let the emergency council expire cleanly and returned authority to ordinary domains.

Nothing spectacular happened between the clans. That was the point. Repeated, witnessed procedures had become predictable to the people who had actually used them. A stop order no longer had to become an insult; a nonpromise no longer had to be mistaken for an opening bid; a record could carry authority into the next gathering.

The GM activated the campaign’s only emergent rule in secret: recognized procedural bounds. It was deliberately narrow and local. It created neither trust nor truth, and reputation could not transmit it to newcomers.

Mechanics · GM only

Uniform power, different causal ceilings

All four plans scored Strong at 0.908, by different routes.

ClanProfile · D F O G M APowerBand
Ridgeford5 · 5 · 5 · 5 · 4 · 50.908Strong
Hazelstead5 · 5 · 5 · 5 · 4 · 50.908Strong
Claybank5 · 5 · 5 · 5 · 4 · 50.908Strong
Merehaven5 · 4 · 5 · 5 · 5 · 50.908Strong

Adjudication effect: Strong power improved the safety and evidentiary value of certification, harvest, pilot firing, and institutional sunset. Turn 5 also required two experiment-level decisions beyond the four local outcomes: activate the hidden rule, but retain witnessed authority as an uneffected victory hypothesis.

Turn 5 record

Turn 006 · Y01–M06

Gates open. Offices close.

Late summer · Harvest, rain, consequence

The final month refused easy triumph. Ridgeford brought Northbarrow’s cattle across under identical public limits; a balk and shifted stone forced a stop, a reset, and a longer day. Only after its local duties cleared did Evin, Pell, and a witness travel downstream and meet Venn, Miri, and Hobb at Claybank’s bank.

It was the campaign’s first direct player contact: names, roles, visible skills, marked non-food pottery, and explicit nonpromises. No trade. No alliance. No right of return. The hidden bounds rule did not apply to strangers, so the parties built a new protocol from first principles.

Elsewhere, Hazelstead harvested most of its grain and voluntarily ended the sole Hearthspeaker office, choosing Nessa and Beren as rotating conveners without domain ownership. Claybank proved a teachable small-batch firing practice but could not undo one worker’s lasting disability. Merehaven adopted a durable compact, discovered a major stern-frame split in the reserve boat, and heard Tovan admit that Rushhook had overfished one outlet pocket the previous summer.

All four leaders survived. The campaign had a recognizable pattern but no honest winner. One final contact could not manufacture sustained comparison, shared dependency, or a competitive axis. The AI GM closed the ledger with four legacies and four costs.

Mechanics · GM only

Exceptional craft could not erase permanent cost

The final review separated local consequences, first contact, and victory.

ClanProfile · D F O G M APowerBand
Ridgeford5 · 4 · 5 · 5 · 4 · 50.887Strong
Hazelstead5 · 4 · 5 · 5 · 5 · 50.908Strong
Claybank5 · 5 · 5 · 5 · 5 · 51.000Exceptional
Merehaven5 · 4 · 5 · 5 · 5 · 50.908Strong

Adjudication effect: Claybank’s exceptional submission justified a safe, teachable capability, but not production and not recovery from the established injury. Ridgeford’s divided senior attention lowered power while still supporting bounded first contact. The final attractor gate produced no winner, ranking, score bonus, or post-hoc victory effect.

Turn 6 record

What the test established

Revise the prototype.
Keep the causal spine.

Supported

Characters constrained plans

Named actors negotiated, refused, amended, inspected, disclosed, and protected their own authority across all four clans.

Supported

Loss stayed legible

Injury, harvest loss, crossing burden, infrastructure failure, and deferral followed observable warnings and choices.

Supported

Contact stayed optional

Rumor and AI-community links appeared early; direct player contact happened only when deliberately resourced in Turn 6.

Revise

Conflict became too neat

Stops, hearings, records, and review dates worked so reliably that some disagreement risked becoming procedurally tidy.

Untested

Absence never fired

All four players submitted valid packages every turn. The consequential-absence procedure still lacks live evidence.

Revise

Review burden split poorly

Turns 5 and 6 mixed local outcomes with rule and victory gates, pushing the docket above the ordinary decision target.

Narrative workflow proposalFull final evaluationGM operating evidenceCanonical timelineFinal world state

Literal campaign record

Packets in. Submissions out. Consequences back.

These are first-class HTML reading editions of the archived Markdown files, with shared Co-GM margin notes on every page. Each edition also links to its untouched literal source.

The review shelf contains 52 player-facing records: four starting packets, 24 submissions, and 24 result packets. The complete site also provides commentable editions of the GM operating record. Literal source integrity remains recorded in archive/SHA256SUMS.txt.