Can competence be earned by living in a digital ecology rather than imposed by an external optimizer?
What substrate do useful AI agents need before autonomy can be trusted?
How should assistants persist, remember, act, and remain legible to the humans around them?
How do agentic systems coordinate real work while preserving visibility and human control?
A non-LLM digital organism where mean intrinsic foraging competence rises through survival and reproduction alone—no external fitness grader.
Tym Huseby
A preregistered, independently audited GHOST study: ecological selection on frozen standing heritable variation enriched genotype distributions with greater inherited capacity for causal, durable, adaptive lived individuality through a preexisting HabitLearner—without inventing the learner, and without claims of consciousness or general intelligence.
Second GHOST paper: ecological selection enriches inherited capacity for causal lived individuality via HabitLearner plasticity—preregistered and independently audited.
Apache-2.0 agent memory with benchmarked recall, MCP tools, and self-hosted Docker.
Self-hosted routing proxy with task classification, fallback chains, and a live dashboard—MIT-licensed on GitHub.
