AI
A benchmark finds personal AI agents work best when matched to specific jobs
A nine-task comparison of Grok Bot, Dot, Muse and Claude Duet found no universal winner: Grok led overall and in several work tasks, Muse was strongest for personal and creative work, and Claude Duet was slower and asked for more approvals. The comparison is limited because the agents did not have identical connectors. A separate Figma practitioner account describes shared AI skills for context gathering, bug testing and growth experiments, while retaining human review at key points.
Editorial summary · Publisher article
What matters
- The benchmark reports Grok Bot as the strongest overall work agent and Muse as the best personal-use agent.
- Unequal access to Slack and GitHub prevented some agents from completing comparable tasks.
- Figma’s PM OS uses shared skills and versioning to standardize routine context gathering and documentation.
- Automated growth experiments retain a human checkpoint before launch, while agents handle analysis and iteration.
Context & caveats
- The benchmark reflects one evaluator’s tasks and setup rather than a controlled product comparison.
- The Figma workflows are a practitioner account, not independent evidence that the approach works equally well elsewhere.
Why this made the edition
The packet contains detailed benchmark observations and a practitioner account of deploying AI workflows, with useful caveats about unequal tools and human checkpoints.
Source material