Key Takeaways

  • Grok Bot won Claire Vo's benchmark with a 93 percent completion rate across nine autonomous tasks, separating itself through Figma manipulation and structured UI documents.
  • Muse captured second place by outperforming the field on personal tasks, creative media, family artifacts, and fast queries, despite lacking work-focused software integrations.
  • Claude Duet scored 79 percent, dragged down by slow execution speeds and a timid safety model that asked for user permission four times compared to once for every rival.
  • Model specialization already beats generalist capability: Claude Duet produced the best code bug fix and launch video, while Grok dominated office workflows.
  • Vo recommends splitting workflows across tools rather than waiting for an all-in-one assistant: route work tasks to Grok Bot and life administration to Muse.

Grok Bot Won Work on Interface Execution

When Claire Vo pit Grok Bot, Dot, Muse, and Claude Duet against nine real-world assignments using autonomous computer control, she was testing whether a model can sit in your operating system and do your job. Grok Bot finished first at 93 percent.

The separation came down to how models handle graphical software. Grok Bot succeeded because it could interact cleanly with Figma, understand canvas coordinates, and generate structured documents inside native applications. While other agents hesitated or misclicked when interacting with dynamic web apps, Grok Bot moved through production software with minimal friction. As Vo summarized: “Here are the results of the very first how I AI personal agent benchmark and Grock one. Grockbot. Grockbot's the best and we love her.”

Claude Stalled on Safety Permissions

Claude Duet finished last at 79 percent. The failure was not raw intelligence. Claude Duet solved the hardest single technical task, fixing a software bug more completely than any competitor and building the highest-quality product launch video.

The issue was operational friction. Claude Duet ran sluggishly and broke autonomy by constantly asking the user for confirmation. “Claude asked for the most permissions. Like Claude was the most timid. So you can see it asked four times for permission where all the other ones only asked once,” Vo noted. In an autonomous computer agent, extreme caution ruins the product. An assistant that stops every two minutes to ask whether it can click a button destroys the productivity gain of delegating in the first place.

The Work and Life Split

The benchmark exposed a hard divide between administrative execution and personal creativity. Muse could not match Grok on office workflows because it lacked integrations into core business tools. But on personal, less structured tasks, it beat the field.

Vo pointed out that Muse dominated creative media and family artifacts: “Muse coming up with all the personal use cases... on the tasks that it had access for, it crushed, especially creative ones.” The takeaway for founders is that the universal personal agent does not exist yet. Running your calendar, household planning, and personal files demands a fast, intuitive model like Muse. Driving spreadsheets, mockups, and tickets demands a system that knows Figma and OS-level file management like Grok.

As Vo put it: “Pick your agent for your job. But if you're picking work jobs, probably give it to Grockbot. And if you're picking life jobs, probably get it to Muse.”

What to Do With This

Audit the computer tasks you plan to automate this week and split them into two buckets: creative personal requests and structured software tasks. Route software edits, Figma updates, and bug inspections to Grok or Claude, but direct personal schedules and family projects to Muse. Stop looking for one agent to operate your entire desktop.