Claire Vo's personal agent bench: Grok Bot first, Claude Duet last
Across nine real tasks, Grok Bot scored 93%, Muse 89%, ChatGPT Dot 88% and Claude Duet 79%, with Claude called the most timid for asking permission four times.
4 AI agents. 9 real tasks. Claude came in dead last. Claire Vo just ran the first How I AI Personal Agent Bench: Grok Bot vs Muse vs ChatGPT Dot vs Claude Duet, same one-line prompts, real jobs from her week. final scores: → Grok Bot 93% → Muse 89% → Dot 88% → Claude Duet 79% her verdict: "Grok Bot's the best." what Grok Bot won outright: → building her show thumbnail in Figma → turning her inbox into a to-do list → shopping for shoes for her it was also the fastest at drafting a Slack reply in her voice and fixing a bug, and one of only two agents that finished all 9 tasks. the part that got me: Claude asked for permission 4 times. every other agent asked once. she called it "the most timid." on a personal agent, every extra "are you sure?" is a task you end up doing yourself. her takeaway, word for word: "if you're picking work jobs, probably give it to Grok Bot." fair caveats: she scored it herself, not blind, and had Claude total up the final numbers from her notes. so it's one power user's week, not a lab benchmark. 99% of people still pick an agent by which logo they already pay for. paste this into Grok Bot: "here are 5 tasks I did by hand this week: [list]. do each one end to end, tell me how long it took and where you got stuck, and show me the result before you send or buy anything." run her 9 tasks on your own stack. I'd bet the ranking holds.
This post quotes another post: “How I AI personal agent bench: 4 agents, 9 tasks”