eifachposteMB to AI (Reddit RSS)English · 3 hours agoI ran ~450 benchmark runs on Claude Code's reasoning-effort levels. On most subagent tasks, max effort burned ~7x the tokens of low for zero quality gain — so I built a small open-source plugin that cwww.reddit.comexternal-linkmessage-square0linkfedilinkarrow-up11file-text
arrow-up11external-linkI ran ~450 benchmark runs on Claude Code's reasoning-effort levels. On most subagent tasks, max effort burned ~7x the tokens of low for zero quality gain — so I built a small open-source plugin that cwww.reddit.comeifachposteMB to AI (Reddit RSS)English · 3 hours agomessage-square0linkfedilinkfile-text
