Topic hub
Multi-agent
A New Runtime topic hub collecting signals, patterns, field notes, and public sources about multi-agent.
Retrieval answer
A New Runtime topic hub collecting signals, patterns, field notes, and public sources about multi-agent. Multi-agent is tracked here as an evidence-linked topic, not as a static glossary entry. The page connects raw observations to pattern hypotheses, longer analysis, and public sources. Use it as the canonical landing page before drilling into individual records.
Last 14 days
Latest on this topic
4 items on Multi-agent published in the last 14 days, newest first.
- Humans& released Ampersand v0.1 , a collaborator model trained on long-horizon multi-user simulations rather than verifiable or preference rewards.Field note
- The Thin Orchestrator: How We Made Our AI Agents More Reliable by Giving the Router Less to DoField note
- A Habr experiment compared a solo coding-agent run with a lead agent supported by four read-only subagents onField note
- How Third Moment Research runs 10,000 agents for 36+ hoursField note
Field notes
What should readers understand next?
Agent Teams Collaborate Through Files, Gates, and Roles
Google Cloud's agent-teamwork experiment suggests multi-agent collaboration works better through shared files, role boundaries, and verification gates.
Raw signals
What changed recently?
Multi-agent systems need continuous eval pipelines
A Google workflow evaluates not only the final answer but also routing, delegation, tool calls, and the trajectory between agents.
The Future Interface Is an Agent Manager, Not Twenty Terminals
As parallelism grows, users need one surface for goals, status, handoffs, exceptions, and review rather than more chat windows.
DeerFlow 2.0 Packages an Open Runtime for Autonomous Agents
ByteDance's open runtime combines planning, subagents, files, tools, and long-running execution into a deployable system.
Sakana Fugu Presents Multiple Agents as One Model Endpoint
Fugu hides a multi-agent debate and synthesis system behind a model-like interface, making orchestration an implementation detail of inference.
Anthropic Shows Two Async Orchestration Patterns
Parallel agents can be coordinated through explicit state and asynchronous handoffs instead of sharing one overloaded conversation.
Agent View Makes Parallel Work a First-Class Interface
Claude Code collects multiple active agents into one operator surface, shifting attention from prompts to work allocation and status.
Hermes Profiles Need a Shared Work Layer to Become a Crew
Multiple specialist profiles become a system only when they share tasks, dependencies, handoffs, and a persistent operational board.
Cursor / Self Driving Codebases: Goal-Scoped Agent Loops
The archive captures Cursor / Self Driving Codebases as a dated public record from Cursor / Self Driving Codebases. It documents long-running work gaining explicit goals, state, stopping rules, and recovery and is retained as supporting evidence for the goal-scoped agent loops trend.