Pick a multi-agent orchestrator from a side-by-side benchmark

The community ran an informal two-iteration shootout of Claude-Code orchestrators on the same product-design task: Ruflo (Claude Flow), Superpowers, Gstack, OhMyClaude, vanilla Claude, Superflow. Use the published ranking as a starting point and run a one-day repeat on your own task before standardizing.

Picking the wrong orchestrator burns a week of setup and learning. The reported rankings:

• Iteration 1 (interface design from a brief): Ruflo > Superpowers > Gstack > Claude
• Iteration 2 (trading-app mockup from voice brief, 2 analyze+fix cycles): Ruflo > OhMyClaude > Claude > Superflow > Superpowers > Gstack

Practical takeaway: Ruflo and OhMyClaude consistently topped; Gstack and Superpowers under-performed on real coding output (Gstack does some marketing — "roasts" your project, recommends YC — but lost on the build).

1. Pick a real, scoped task you'd ship this week (a small product mockup or a feature spec).
2. Install the top two from the reported ranking: Ruflo and OhMyClaude (oh-my-claudecode + oh-my-codex bundle).
3. Run the same identical prompt through each.
4. Score on: did it identify the killer feature? Token usage? Iterations to converge? Bug rate? Aesthetic of output?
5. Pick a winner for your task class; stop swapping.
6. Re-benchmark quarterly — the orchestrator landscape moves fast.

• The published ranking is subjective ("how I felt about the result"), not a rigorous matrix. Treat it as a starting bias, not a verdict. Your task class matters more than the global ranking.