The repository says the simulation spans 45 eras, from prehistory toward speculative futures, and that events are not scripted. The player can observe the world or intervene in it.
This starts a distinct analytical record because no prior New Runtime record is supplied for TheHumanBox; the relevant question here is the boundary between explicit simulation rules and model-generated behavior.
Thesis
TheHumanBox makes a civilization simulation inspectable as a local, rule-driven software artifact rather than a hosted AI service.
Observed facts
The repository describes autonomous inhabitants forming tribes, developing languages, building cities, founding religions, fighting wars, signing treaties, and passing culture across generations.
1The repository says the simulation spans 45 eras and is written in Rust compiled to WebAssembly, runs on the user's device without a server or LLM, and saves worlds locally for export.
1
Mechanism
The Rust simulation engine compiles to WebAssembly and runs on the user's device; the repository says there is no server or LLM in the browser simulation, and that worlds save locally and can be exported.
Why now
A project promising autonomous social emergence is easier to evaluate when its execution boundary and source are visible, rather than hidden behind a hosted model.
Implications
Builders can inspect and modify a simulation whose outcomes are generated by explicit rules, not model calls.
Evidence boundary
These facts come from the project's GitHub README and repository description; they do not independently validate behavioral realism, the quality of its emergent outcomes, or the full feature claims.
Watch conditions
Whether independent users can reproduce interesting, non-scripted outcomes from the published code and explain which rules produced them.