At a glance
WHAT IT’S REALLY ABOUT
Claire benchmarks four personal AI agents across nine real tasks
- Claire Vo runs a nine-task “How I AI” personal-agent benchmark across Dot, Grok Bot, Muse, and Claude Duet using short, one-line prompts and an operator-agent workflow.
- She observes major differences in speed, permissions behavior, and tool access, including that Muse lacked business connectors and Claude Duet was slow and asked for the most permissions.
- On individual tasks, Grok stood out in Figma thumbnail creation, inbox-to-do formatting, and shoe-shopping/price comparison, while Muse shined on creative personal artifacts like a family newspaper and fast homework explanations.
- Coding results were mixed: Dot was closest to the main fix, Claude Duet found an additional bug and avoided unnecessary changes, and Grok’s approach was weakest by loosening validation.
- In the final scoring run through Claude, Grok wins overall, with Claire concluding Grok is best for “work jobs,” Muse for “life jobs,” and that consistent access plus good UX artifacts are critical in evaluating agents.
IDEAS WORTH REMEMBERING
5 ideasThe benchmark aimed for fairness, but still had real-world setup constraints.
She tried to keep things comparable by choosing cross-cutting tasks suggested by the agents themselves, standardizing tool access where possible, and using one-line, “in my voice” prompts with an operator agent driving execution.
Grok Bot was the overall best performer in Claire’s scoring.
Grok completed all tasks and was ultimately scored as the overall winner; Claire says it won on breadth, taking top spots on thumbnails, inbox to-dos, and shoe shopping.
Muse excelled at fast, personal/creative tasks when access wasn’t a blocker.
Muse could only do seven of nine tasks due to missing work connectors, yet Claire says it “crushed” the tasks it could access—especially creative outputs, the family newspaper, short-form content design, and fast Q&A-style explanations.
Dot came off as competent but weak on creative execution and polish.
Claire says Dot did well on some practical tasks (including drafting a Slack reply that matched her preferences and doing solid coding), but repeatedly failed at creative artifacts like the family newspaper and the launch video.
Claude Duet was uneven: strongest on deeper coding/video walkthrough quality, weakest on speed and task completion.
Claire found Duet slow and overly permission-seeking, and it failed some tasks (notably short-form video editing), but she says it fixed the bug most completely and produced the best Product Intelligence launch video with strong branding and value prop.
WORDS WORTH SAVING
5 quotesI told you all I have personal-agent fatigue, but I am here for you.
— Claire Vo
One is, man, I'm now gonna have computer use run all my agents running other agents. Like, the parallelism of this exercise was Thrilling to the heart.
— Claire Vo
Claude computer use is just hot garbage.
— Claire Vo
If you're using Figma, use Grok Bot, man. It is so good.
— Claire Vo
Pick your agent for your job, but if you're picking work jobs, probably give it to Grok Bot, and if you're picking life jobs, probably give it to Muse.
— Claire Vo
High quality AI-generated summary created from speaker-labeled transcript.
