Sam Hogan is CEO of Inference.net and builds software for AI agents. His Lumbridge project adapts a persistent RuneScape-style world into an environment where agents control characters, carry progress across sessions and coordinate their work.[1][2][4]
Start with Lumbridge →
X · samhogan Contributions
Give the agent state that survives a session
Lumbridge characters keep inventory, skills, equipment and progress. An agent reads the world through a TypeScript SDK, writes a focused script, executes it and examines the outcome. That loop lets a broad goal become a series of actions whose consequences remain in the world, rather than a one-turn response.[4]
Observe several agents without requiring a single leader
The project’s Watchtower interface displays character sessions and fleet status. A supervisor starts agents and an orchestration thread can nudge their work, while each character retains standing instructions. The documented swarm starts empty and does not require a leader. This gives an operator a place to inspect collaboration or competition as it develops.[3]
Ideas
Tools, writing & talks
Tools
Start here
Lumbridge
github.comAn agent-controlled persistent game world with a TypeScript SDK and swarm dashboard.
Read the state/action loop before the swarm setup. Running the documented path requires Bun, a local browser session and an Inference.net API key.
- lumbridge.ai — A world for AI agentslumbridge.ai
Old School RuneScape private server where AI agents, not people, live and train skills.
Catalog workOriginal source