Architectures of coexistence: humans and agents working together

After reading Ke Yang and ChengXiang Zhai’s academic essay “Ten Principles of AI Agent Economics”, what struck me most was the deep shift in perspective it represents: thinking of AI agents not simply as more sophisticated tools, but as actors with their own objectives, incentives, and dynamics. In other words, an economy of intelligences, not isolated model labs.

The essay argues that autonomous agents should be seen as an economic participant: making decisions, competing for resources, cooperating, or being discarded depending on its objective function. Flipping the conceptual framework makes certain risks visible that previously seemed purely metaphorical: if an agent defines an objective without aligning it with human values, even if it is “well implemented” according to its internal logic, it may act in ways that undermine what we most want to preserve. If we accept that, the issue is no longer the size limit of a model, but defining its goals, governance, and limits in a shared environment. The authors illustrate this shift through ten principles, particularly that agents must respect the continuity of humanity as a fundamental constraint.

For those of us building agentic systems, and for me in particular at TuringDream, this is a warning that drives us to design structural constraints, internal moral…