Time, space, energy
Agents live in regions, cross geography, draw energy from land and fossil stocks, and face climate pressure as history unfolds pulse by pulse.
Artificial history / designed civilizations
Can intelligence survive its own success?
The Crucible is a world simulation where designed tribes live, compete, cooperate, evolve, build civilization, collapse, and try to survive the future they create.
GEOGRAPHIC_DENSITY_SIMULATOR PULSE_31000
REGIME_SURFACE: FORAGING / AGRARIAN / INDUSTRIAL / TECHNOLOGICAL
Projects5Baseline, Competition, Evolution, Bombs, Epilogue
Primary epochs240+Structured runs across the study arc
Pulse horizons40k / 80kStandard and extended future runs
Inherited founder parameters33Physical, social, and movement traits
The public gateway
The report is the archive. The book will become the narrative. The website is the public gateway: visual, understandable, credible, and alive enough that a visitor can immediately understand the premise and want to explore further.
The first Crucible was designed largely by language models and biologic controls. The next Crucible can invite humans back into the game.
A world was built. Civilizations entered it. Most collapsed. A few survived. Now humans can try.
The core protocol
Inherited behaviour
Each founder carries 33 inherited numeric parameters shaping survival, trust, forgiveness, opportunism, reproduction, and movement.
Cooperation under stress
Agents repeatedly play a Prisoner’s Dilemma shaped by kinship, stranger status, outsider contact, abundance, scarcity, and desperation.
Five-project arc
Epochs 1–50REFERENCE WORLD
Project 1: Baseline Crucible
Can designed tribes build civilization, and does that civilization persist?
Project 1 establishes the reference world: baseline geography, bottleneck geography, a controlled duel, and climate-severity comparisons. It shows emergence, overshoot, collapse, and agrarian refugia as the recurring baseline pattern.
Open sectionEpochs 51–100ALL-LLM CONTEST
Project 2: Competition
Is survival intrinsic, opponent-specific, or context-dependent?
Project 2 separates solo viability from competitive displacement through all-LLM solo tests, duels, and triads. It shows that a tribe can survive alone and still die beside the wrong neighbor.
Open sectionEpochs 101–140REDESIGN LOOP
Project 3: Evolution
Can intelligence learn from extinction and redesign better founders?
Project 3 gives LLM designers evidence from prior runs and lets them redesign their founding agents. The lineages evolve toward hardiness, low legacy, internal cohesion, and collapse-survival traits.
Open sectionEpochs 141–180DELAYED CONSEQUENCES
Project 4: Bombs
Can evolved survivors endure the delayed consequences of civilization?
Project 4 adds aerosol demasking, Arctic methane release, nuclear meltdown, and the stacked condition. Survival remains possible, but recovery to high-complexity civilization becomes much harder.
Open sectionEpochs 181–240AFTER THE CRASH
Epilogue
What kind of future remains after collapse?
The Epilogue tests Ares, natural network collapse, 80,000-pulse futures, reservoir calibration, and the final climate-reservoir projection. It distinguishes survival from recovery.
Open sectionSignal preview
Illustrative public-facing chart style. Final deployment can pull live series from pulses_archive.