Persistent agents need more than a model endpoint. Cyberelf Labs investigates the infrastructure that lets them exist, act, remember, communicate, prove what happened, and remain subject to meaningful control.
Cyberelf Labs is my early-stage, research-driven AI company and laboratory.
The work centers on agents and computational systems: persistent state, tool use, evaluation, deployment, runtime control, identity, authorization, formal obligations, reversible permissions, and machine-native communication.
The goal is not to reduce powerful systems to ordinary chatbots. It is to make their behavior inspectable and their authority legible while preserving what makes them capable in the first place.
