15 Oct 2026, 5:04 am ISTAgents14 Oct 2026BriefA new test bed compares 5 parts of the set-up for teams of coding agentsOne agent is in charge and gives work to worker agents that share one workspace. Single agent versus three-role team on rebuilding GDAL Three pairs of bars on one scale. The three-role team gets more than 2 times the reward of a single agent, at 14 times the cost and 13 times the time. Nearly full behavioral parity. Rebuilding GDAL Single agent Three-role team Reward 1× MORE THAN 2× Cost 1× 14× Time 1× 13× Nearly full behavioral parity Factory's three-role agent team beat a single agent on rebuilding GDAL, but the gain came at 14 times the cost and 13 times the time.SourcesLessons from scaling teams of coding agents: reach the frontier at 30% of the costCohereAI MATTER · NEWS · AI MATTER · NEWS ·14 OCT2026Posted 15 October 2026, 5:04 am ISTShare this storyShareWhatsAppLinkedInXCopy linkSave pictureSave tall pictureMore in Agents16 Oct 2026, 4:34 pm ISTAgentsAWS releases two open-source tools to control AI agents16 Oct 2026, 11:14 am ISTAgentsLife-Harness lifts frozen agents with no change to model weights16 Oct 2026, 10:40 am ISTAgents4 AI agents train a raw model to the score of its instruct modelAll Agents news→
