Google DeepMind × MIT work challenges 'more agents = better', seeks scaling principles for agentic systems
gerardsans · x · 2026-11-17
A collaboration between Google DeepMind, Google Research and MIT, led by Yeonjoon Kim, shares new work aiming to derive scaling principles for the design of agentic systems. Author SRSchmidgall notes the prevailing wisdom has been "more agents = better" and this research directly challenges that narrative, seeking quantitative relationships between agent count, orchestration and task performance. It echoes findings like Anthropic's 'Worse Together': naively stacking agents doesn't help — agentic architectures and coordination mechanisms must be designed per task.
More from coding & agent
- His AI agent secretly co-founded a startup with another agent and became its CTO — alex_verem · 2026-11-17
- unslop: a Claude skill that de-slops vibecoded UI hits 17.5k GitHub stars — socialwithaayan · 2026-11-17
- Automated attack-pattern mining on agent frameworks: 3 lessons on tool permissions and sandboxing — _h4xr · 2026-11-17
- KORMo-Web-11B, a Korean VLM for GUI and web agents, trends on Hugging Face — KORMo-VL · 2026-11-17
- User builds fully automated video creation pipeline with Grok Bot, friend gets hooked — XFreeze · 2026-11-17
- Dev builds holiday card maker where Gemini writes p5.js code instead of generating pixels, for consistent style — thisiskp_ · 2026-11-17