Map/Agents and organisations/Wave 9

Multi-agent coordination

Agents dividing work, sharing context, handing off without loss and catching each other’s errors, so a team of agents beats its best member.

capabilityEmerging Open on the map →

Facts

Lane: Agents and organisations
Wave: 9 of 12 (build order, not a date)
Needs: 14 boxes all the way back through hard links
Unlocks: 2 boxes downstream through hard links
Leverage: 0.2 points · stands in the way of 2 of 53 stages · rank 45 of 57
Cluster: Trust plumbing and the agent economy

Evidence as of October 2026

Google Research (Jan 2026) found that, across 180 agent setups, every multi-agent variant made sequential planning 39-70% worse, and independent agents amplified errors 17.2x.

Around it

Needs · 5

Long-horizon agentsstage · Built today
Context engineeringcapability · Built today
Agent-to-agent protocolscapability · Emerging · helps, not required
Persistent memorycapability · Emerging · helps, not required
Institutions for agent societiescapability · Speculative · helps, not required

Unlocks · 4

Agent economiesstage · Speculative
Explosive economic growthstage · Speculative
Agent-native organisationsstage · Emerging · helps, not required
Drop-in remote workerstage · Speculative · helps, not required

Brakes

Works against this box

Misbehaviour cascades in agent swarmsbrake · Emerging

Loops and off-ramps

Capex pullbackoff-ramp · cut off
Taiwan chip disruptionoff-ramp · cut off
Power shortfalloff-ramp · cut off
Capability plateauoff-ramp · cut off
Training data ruled outoff-ramp · slowed

Openings researchers' judgement, October 2026 scouting

Startup room 35/100 · rank 90 of 124 Research room 54/100 · rank 49 of 124

Crowding3.5 Lab overlap4.5 Small-team fit3 Buyer today3 Research openness4.5

Depth: light
Players: Anthropic Agent Teams (Opus 4.6); Google A2A (150+ orgs); Isara ($94M, $650M val, OpenAI-backed, Mar 2026); Sycamore ($65M seed); BAND ($17M seed); Lua ($5.8M); Modern Relay ($3M seed)
Wedge: Research on when a team beats its best agent: handoff and context-transfer protocols plus verifier designs that contain error cascades on sequential tasks, packaged as a topology selector that existing orchestrators can call. research