Teams of Agents

Run the workers one at a time and the lead waits on the slowest one and cannot steer any of them mid-task. Run them at once and you take on keeping their results, their state and their errors consistent. Nobody has both.

What people are trying

Anthropic runs subagents one at a time today because it is the easier half to coordinate, and treats concurrent execution as worth having once the coordination problems are handled rather than as something already solved.

• How we built our multi-agent research system · Anthropic · read 09/19/2026

  But this asynchronicity adds challenges in result coordination, state consistency, and error propagation across the subagents.

  Dated June 2025. It is still the maker's own account of this architecture, and nothing newer from them supersedes it.

Where it bites: Lead agent and workers

When one agent hands a task to another, the authorization to act on it often travels with the task. Passed in band, that credential is visible to every agent in the chain and not only to the one it was meant for.

What people are trying

The Agent2Agent specification asks that a credential be bound to the agent that originated the request, and encrypted when it carries anything sensitive, so only that agent can use it. Its own preference is to deliver credentials out of band rather than through the chain at all.

• Agent2Agent (A2A) Protocol Specification · Agent2Agent Protocol Project, Linux Foundation · read 09/19/2026

  In-band credential exchange can allow credentials to be passed across chains of multiple A2A agents, exposing those credentials to each agent participating in the chain.

Where it bites: Agent graphs

Asking several models to review the same work looks like several opinions and is not. Their mistakes are correlated, so a panel carries a fraction of the independent judgment its size suggests, and one measurement found nine judges worth about two independent votes.

What people are trying

Measuring a panel against what genuinely independent voting would give, rather than assuming more reviewers means more reliability, and testing whether cleverer ways of combining votes recover the difference. In that measurement they mostly did not.

• Nine Judges, Two Effective Votes: Correlated Errors Undermine LLM Evaluation Panels · arXiv · read 09/19/2026

  the 9 judges effectively provide only about 2 independent votes' worth of information.

• Correlated Errors in Large Language Models · arXiv · read 09/19/2026

  We find substantial correlation in model errors -- on one leaderboard dataset, models agree 60% of the time when both models err.

Where it bites: Review and debate