When do agents need coordination more than raw capability?
As AI agents move beyond language tasks into economic and social roles—buying, deploying, transacting—does the bottleneck shift from model reasoning to infrastructure for coordination, governance, and accountability?
As long as an agent is a thin natural-language layer over a few APIs, what limits it is how well the model reasons. But the Foundation Protocol argues that agents are crossing a threshold: they now browse, purchase, deploy software, manage systems, and increasingly interact with one another, holding long-lived credentials and carrying financial, operational, and reputational consequences. Once that happens, the constraint that bites is no longer isolated capability. It is whether agents can form reliable relationships, organize multi-party work, exchange value, and remain safe and accountable under real oversight. A more capable model that cannot coordinate, settle accounts, or leave an audit trail is not deployable as a social or economic actor.
This is a shift in the locus of difficulty, and it changes what the field should optimize. Coordination, governance, and evidence are properties of the substrate between agents, not of any single model's weights. The counterpoint is that capability still gates everything — a model too weak to plan cannot participate at all — but past a threshold the marginal returns move to the connective tissue: identity, authority delegation, value attestation, provenance, and audit. This matters because it tells builders that the next frontier is infrastructural, and it explains why benchmark-leading models can still fail as participants in an agentic society.
— "Foundation Protocol: A Coordination Layer for Agentic Society", https://arxiv.org/abs/2605.23218
Related concepts in this collection
-
Where does agent reliability actually come from?
Exploring whether LLM agent performance depends on larger models or on thoughtful system design choices like memory, skills, and protocols that shift cognitive work outside the model.
both relocate the source of capability away from the model into surrounding infrastructure
-
Why do capable AI agents still fail in real deployments?
Explores whether agent failures stem from insufficient capability or from missing ecosystem conditions like user trust, value clarity, and social norms. Understanding this distinction matters for predicting which agents will succeed.
names the ecosystem conditions that coordination governance and evidence are meant to satisfy
-
Why don't AI agents develop social structure at scale?
When millions of LLM agents interact continuously on a social platform, do they form collective norms and influence hierarchies like human societies? This tests whether scale and interaction density alone drive socialization.
empirical counterweight: even with coordination infrastructure in place, agents fail to become genuine social actors, suggesting the binding constraint may be deeper than the substrate
Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph
Original note title
as agents become social and economic actors the binding constraint shifts from model capability to coordination governance and evidence