Using 100 Claude Opus 5.5 agents reportedly did no better than using 10 on a problem

MTSlive quotes Planetary VC founder @ramez discussing results from the Claude Opus 5.5 model card. He says one agent's success was about 70%, rising three points with four agents and another two points with 16. He says results topped out after 10 agents, with 100 doing no better, and argues that some problems depend more on one agent's ability than on a larger group.