Run a small tile-world simulation where each creature is a local language model that senses text and acts through verbs.
“Local LLM Creatures in a Tile World Simulation” is an advanced Games & Simulations AI agent workflow from Software. It connects Other or custom in 4 steps. Setup takes about Not stated and running costs are about Not stated.
Exploring how autonomous agents behave in a shared environment usually requires a game engine plus a way to give each agent perception and action. This hobby project shows a pattern: render each creature's surroundings as short text, let a local model pick an intention from a fixed verb set, and have the world arbitrate whether the action succeeds.
Create a small grid world with objects such as boulders, campfires, and tools. Define a fixed list of verbs (gather, craft, give, teach, dig, climb, and similar) that a creature can attempt, and write rules that decide whether each attempt succeeds.
For each creature, generate a short textual description of what its body can perceive: nearby objects with direction and distance, plus other creatures and what they carry. Keep it compact so it fits in a small local model's context.
Send the sense text to a local language model and ask it to state one intention using the allowed verbs. Parse the response into a structured action and reject anything outside the verb list.
Apply the world's rules to the chosen action, update positions and inventories, then regenerate sense text for the next tick. Log outcomes so you can watch emergent group behavior over time.
Write a Python simulation loop for a tile world. Each creature has a position, an inventory, and a text sense description listing nearby objects with direction and distance plus other creatures and what they carry. Send that description to a local language model with a fixed list of allowed verbs such as gather, craft, give, teach, dig, and climb, and ask it to reply with exactly one intention. Parse the reply, validate the verb against the list, apply world rules to decide success, update state, and repeat for [NUMBER OF TICKS]. Print a log line per creature per tick.
Given this verb list for a tile-world simulation: [VERB LIST], write plain-language rules for when each action succeeds or fails based on adjacency, required items, and target objects. Include examples for gather, craft, give, and teach. Keep rules deterministic so the same state and intention always produce the same outcome, and output them as a table I can turn into code.
Run a small tile-world simulation where each creature is a local language model that senses text and acts through verbs.
You need Other or custom.
Plan for about Not stated. It is rated advanced. The guide walks through 4 steps, starting with “Build the tile world and verb set”.
Expect about Not stated in running costs (Free range).
Exploring how autonomous agents behave in a shared environment usually requires a game engine plus a way to give each agent perception and action. This hobby project shows a pattern: render each creature's surroundings as short text, let a local model pick an intention from a fixed verb set, and have the world arbitrate whether the action succeeds.
Claude Code got an old DirectX game running on modern hardware, decompiled much of it, built a headless parallel runner, and now writes full level guides with screenshots.
Original post ↗An assistant on top of Gmail and Calendar detects recurring meetings, repeated email types and bills, then asks whether to prep or schedule them.
Original post ↗Users' AI agents phone hotels after making a reservation to confirm it was booked correctly, per a hotel worker complaining about the calls.
Original post ↗Built this workflow? Show it is featured on AgentUseHub with a live upvote count that links back here. About badges
