Back to Transient Labs

Equilibrium Design

We design the game.
Not just the agents.

Every production multi-agent system is a game — agents with objectives, constraints, and incentives. We architect the equilibrium so the system converges to the right outcome by construction.

Nash equilibria, Stackelberg hierarchies, and observable payoffs are not academic abstractions. They are the engineering primitives behind agent systems that hold up under pressure.

L0Payoff Matrix
The Rational Agent

Change the payoffs. Watch the equilibrium shift.

B: Cooperate
B: Defect
A: Cooperate
A
3
B
3
A
0
B
5
A: Defect
A
5
B
0
Nash
A
1
B
1

Both defect — the system converges to mutual loss. Redesign the payoffs.

L3Budget Wars
Intra-Firm Dynamics

Switch allocation policies. See who games the system.

Requested
Allocated
Engineering38% / 92%
Marketing35% / 85%
Sales32% / 78%
Ops36% / 88%

Proportional allocation creates inflation incentives. Agents learn to overstate need — the system diverges from truth.

L5Trust Game
Human x AI Equilibria

Toggle observability. Watch trust compound or collapse.

Observability
H
Human
50%
AI
Agent
CollapseStable

Observable payoffs compound trust. Both sides converge to cooperation.

L6Stackelberg Hierarchy
The Autonomous Enterprise

Expand each layer. See who governs whom.

Stackelberg hierarchyClick to expand

Each layer sets policy before the layers below optimize. The architecture is incentive-compatible by construction.