promerget is an independent research practice exploring how agent-based AI architectures can be designed, coordinated, and made reliable in complex knowledge work.
Our work follows a three-phase cycle. We design controlled experiments around specific questions, run them in isolated environments, and document what we learn. Outputs are research artifacts — not services.
We start with a narrow, falsifiable question about how agentic systems behave under specific conditions.
We build minimal agent architectures targeted to the question and run controlled experiments in isolated environments.
We analyze the data, reconcile it with the hypothesis, and document findings — including null results.