Implicator PRO Briefing / 12 Nov 2026
|
The Decision Brief If your team runs an AI agent with a router or orchestrator in front of specialist subagents, Anthropic now says most of those subagents should be folded into one agent loop. Its Oct. 22 guidance says one agent with skills frequently beat subagent designs on quality, often at lower cost and latency, across enterprise deployments it does not name. We examined the outside evidence that has appeared since: Vals AI's test of single agents against agent teams on 50 web apps, an independent run of Anthropic's own migration kit, a Google Research study of 180 agent setups, and Nous Research's code cleanup with 1,393 subagents. In the Vals test, teams cost 1.8 to 5.1 times as much as single agents, and only one of four comparisons showed a statistically significant quality gain. None of these sources tests the deployments Anthropic cites, and we ran no tests ourselves. The full piece shows which kinds of work favored teams in these tests, and sets out a proposed fold-or-keep record for auditing one subagent at a time. The evidence and the fold-or-keep record are below for PRO members. View PRO plans, $10 monthly or $99 annually. |
Marcus Schuler
Editor-in-Chief and founder of Implicator.ai. Former ARD correspondent and senior broadcast journalist with 10+ years covering tech. Writes daily briefings on policy and market developments. Based in San Francisco. E-mail: editor@implicator.ai
IMPLICATOR