Summary

Matt Rocklin shares his initial experiments with multi-agent AI workflows using his Claude TUI alternative, Claude Chic. He ran two experiments—a Diplomacy game between agents and a collaborative session viewer build—finding that coordination overhead often outweighs parallelism benefits. He concludes that simple multi-agent patterns like separate review and fresh perspective agents are clearly valuable, while fully parallel 'YOLO swarms' are not yet worth the coordination cost. His key insight is that developers should focus on what's actually holding them back rather than chasing multi-agent hype, and that sequential agent workflows often beat parallel ones.

Key Insight

Multi-agent AI workflows are most valuable in simple sequential patterns like review and fresh perspective, while fully parallel swarms suffer from the same coordination overhead that plagues human teams.

Spicy Quotes (click to share)

• 5

  No one is behind because no one knows what they're doing yet.

• 2

  Multiple coordinating agents are capable of more sophisticated behavior than single agents.

• 6

  The solution to AI problems is more AI.

• 5

  Also like humans, sometimes teams don't perform as well as single developers.

• 3

  There's a real cost to coordination that needs to be outweighed by the benefits of speed.

• 5

  AI agents are so fast that it's not clear to me when we'd want multiple agents working concurrently on the same code—most tasks finish before coordination overhead would pay off.

• 6

  Rather than ask 'How can I use multi-agent swarms?' I am instead asking 'What is holding me back right now? How can I best address that?'

• 3

  Regardless what you think about AI, and regardless how you think about full-auto multi-agent swarms, you should experiment with them.

Tone

reflective, pragmatic, exploratory