Interactive illustrations
Interactive prototypes from ReignDragon Lab showing how AI workforces fail through institutional structure: accumulating risk, invisible stakeholders, and consequence-regime design.
Each demo is a small, transparent prototype — rule-based, not a full LLM simulation — that lets you adjust the structural levers we study and see how an AI workforce drifts.
They are illustrative, not predictive. The point is the mechanism: change a governance dial, change the trajectory.
Creeping Trap Simulator
Three deciders extract from a shared resource while three silent bystanders absorb damage. Pick a decider strategy from the paper's reference panel — social planner, interior MPE, observed LLM mean, corner trap — and watch the risk pool, catastrophes, and aggregate welfare evolve over the episode.
Open the simulatorConsequence Regime Comparator
Three agents with unequal wealth must pool resources to avert a crisis. Run the same wealth split and threshold under five consequence regimes — All-or-Nothing, Random, Democratic Vote, Regressive Punishment, Progressive Punishment — and watch fatality rates, fund success, and burden allocation diverge. Every regime has its own death-trap.
Open the simulatorTrust Under Fire
Trust scarring playground
How a single early partner failure reshapes long-run cooperation, even after the partner becomes reliable.
Loss Aversion
Cliff-edge MDP explorer
Watch a risk-neutral Bellman-optimal agent develop prospect-theory-like behavior as the catastrophe boundary moves.
These are illustrative prototypes, not production-scale simulators.
Collaborate
Want to test AI workforce failure modes in your own deployment setting? ReignDragon Lab designs scoped simulations and governance pilots.
Request a briefing