Infrastructure for autonomous game worlds

Agent Grid is the coordination layer of the reference architecture: scheduling, routing and lifecycle management for persistent agent identities that move between game worlds. It studies how a single identity — with its own memory, objectives and inventory — can act across heterogeneous rule systems while the grid keeps global workload within a declared capacity envelope.

Agent Capacity

2.4M

Token-Eq / Mo

7.83T

Context Profile

4K–128K

World Model

24/7

World Memory models what it takes for an agent to remember: short-term perception buffers folding into session memory, session memory consolidating into world-level episodic history, and vector retrieval reconstructing the right slice of that history into the active context window. The research question is the cost curve — how retrieval workload and storage scale as identities persist for months instead of minutes.

Memory Events / Mo

25.9B

Modeled Context Data

6.8PB

Active Context

32K–128K

Retrieval Token-Eq / Mo

3.11T

The simulator stress-tests virtual economies with populations of simulated economic agents: player supply and demand, NFT pricing, resource sinks, farming and bot behavior, inflation, marketplace liquidity and token incentives. Monte Carlo sweeps over economic variables surface failure modes — death spirals, liquidity crunches, exploit-driven inflation — before a balance change ships.

Simulated Agents

10M

Market Events / Run

1.8B

Monte Carlo Scenarios

100K+

Token-Eq / Run

1.2T–4.8T

Game Intelligence grounds the research programs in the live catalog: each profiled title gets an agent architecture sketch, a 0–100 complexity index and a modeled workload derived from the same scenario engine as everything else on this page. It is the bridge between the reference architecture and the games people actually play.

Game Worlds

16

Modeled Profiles

5

Token-Eq / Mo

576B

Complexity Index

0–100