The Long Watch

The game runs from the 2030s into the 2070s in 8 eras. In each era you are offered three of the 18 things that could happen next — an open release of a powerful model, a treaty on computing power, a collapse of trust in what is real, a guarantee of care that no ruling can close. You decide which one happens, whether it arrives early or late, and whether it is done openly and accountably or in secret.

Published rules then move seven measures you can watch the whole time: capability, concentration, dependence, checkable truth, public voice, resilience, care. Between eras the measures act on each other — capability pulls dependence up, power that cannot be checked spends care — so a run can be lost slowly without a single decision that looked wrong at the time. Every run ends in one of 8 named endings with a plain account of why.

• A finished run is stored as its seed and its choices, printed as a short code.
• Anyone holding that code, or the link a finished run becomes, replays it against the published rules and reaches the same ending.
• Return to any era in a finished run, try another choice and compare the paths from the same starting conditions. A comparison link keeps both runs checkable.
• Nothing about a player is recorded — there is no account, no name and no free text anywhere in the game.

The public ledger, its statistics and the weekly shared world are the next thing to build, and they are not open. Until they are, a run lives in your own browser and is shared only if you choose to pass the code to someone. The privacy page states exactly what that means today and what will change when the ledger opens.

The Long Watch is a rule model of institutions and technology. It makes no spiritual claims, computes no real-world probabilities and names no real dates, people or organizations. My Christian worldview is argued separately, in the paper — and the paper’s own scenario has an explorer of its own, which is a different thing entirely.