The work
One game master prepared a campaign intended to be played with other people. AI tools were part of that creative workflow. The author chose to develop it, made editorial decisions, and invested real time in doing so.
There was supposed to be a tabletop game. There were hours of preparation. There were friends to invite. And then there was a disagreement about the tools used to make it.
This is the archive of Message in a Bottle, a planned Savage Worlds campaign. The intended destination is a full, unapologetically spoiler-filled account of the adventure that could have unfolded.
One game master prepared a campaign intended to be played with other people. AI tools were part of that creative workflow. The author chose to develop it, made editorial decisions, and invested real time in doing so.
Because the game was not played, there is no canonical session log. No real player choices, discoveries, dice results, or endings exist to report. Planned paths must be described as possibilities, never as events.
The final page will walk through the GM's actual plan, including the opening situation, pressure points, hidden information, possible decisions, and proposed endings. The source text for those parts is still needed.
What to insert: the real premise, character hooks, first scene, and intended first choice. Keep the difference between player-facing setup and GM-only facts explicit.
What to insert: the actual NPC agendas, secrets, threats, clues, and rules or mechanics that mattered to the scenario. No invented antagonists.
What to insert: alternate approaches, failures that move the story forward, and consequences that depend on player choices. Avoid presenting branching prep as a predetermined plot.
What to insert: the planned revelation and the evidence designed to make it land. This section should contain the real spoilers once the original notes are available.
What to insert: the prepared outcome paths, their triggers and costs, plus any unresolved finale design. Not a single invented “canon” ending.
This page does not identify or shame anyone who declined to participate. The point is to document a piece of human-directed creative labor that might otherwise be reduced to one adjective: AI-assisted.
Even a complete set of GM notes is not a replay of the game. A tabletop campaign is a blueprint for a collaborative event, and some of its best moments can only happen when other people arrive and do something unexpected.
The title, system, approximate time spent preparing, the involvement of AI tools, and the fact that play did not occur.
The source-verified plot, named characters, encounters, branches, reveals, and endings. Those should be added from the GM's original materials before release.
No. The adventure was prepared, not played. Even after source notes are recovered, this page will describe what the GM planned, not pretend that a session occurred.
Its human author developed it using AI-assisted tools. The publication should show the author's intention and decisions, and be candid about AI involvement.
Because “what would have happened” is an archival claim. A convincing invented ending would be fiction about the campaign, not evidence of the campaign that was built.
It was a disagreement about whether the tools used to make the work made it unacceptable in advance. Those are different questions.
Read “You Don't Get to Choose My Tools” →Austen Tucker · The Arcades' Lab · Draft archival presentation