Same Task, Two Bots, Same Time: Why We Ran This Experiment
One day, curiosity struck. 「What if we send the same data to two Claude Code bots simultaneously?」 Would it be faster than one bot? Or would things get messy? The dream team (Hamsters + Puppies) started a casual experiment in our daily workflow.
Our team operates using a Windows device + Mac Mini combo. So the question became: 「If we run the same task on 2 devices at the same time, will efficiency actually double?
Experiment Setup: Keeping It Simple
The test subject was a data cleaning task.
**Task**: Spreadsheet data (100 rows) → Remove duplicates + Standardize format + Sort by priority
Conditions:
• Bot A (Hamsters): Claude Code running on Windows device
• Bot B (Puppies): Claude Code running on Mac Mini
• Start simultaneously, use the same source file
• Save results in separate folders
What Actually Happened
1️⃣ Speed: Not As Fast As We Expected
Expectation: 「It'll be twice as fast」
Reality: 「No significant difference」
• Bot A completed: 3 min 42 sec
• Bot B completed: 3 min 38 sec
The Claude API response time was the bottleneck. Even though both bots sent requests simultaneously, the server still seemed to process them sequentially.
2️⃣ Accuracy: Consistency Was Surprisingly High
When we compared the two bots' output:
✅ Data cleaning results: 99% identical
✅ Sort order: Perfectly matched
❓ Detail formatting: 1 minor difference
The subtle difference was just one date format. Our prompt didn't explicitly say 「YYYY-MM-DD format」. Bot A interpreted it as 「2024-01-15」, while Bot B went with 「01/15/2024」.
Key Takeaway: Even with identical prompts, interpretation can vary slightly. You need to specify even hyphen-level details explicitly.
3️⃣ Resources: How Much Device Load?
Windows device (Bot A): 45% CPU usage during execution
Mac Mini (Bot B): 38% CPU usage during execution
We confirmed that 「simultaneous execution doesn't increase the load on each device individually」. Claude Code is lighter than we thought.
Can We Actually Use This in Real Work?
✅ Where It Makes Sense
1. **Independent Batch Processing**: 100 customer records split into 10 groups, 10 bots processing simultaneously
2. **Multi-language Translation**: Generate Korean, Chinese, and English versions at the same time
3. **Validation Work**: 2 bots independently review the same data, then compare results
❌ Where It Doesn't Work
1. **Sequential Dependency**: 「Bot A's result → Bot B does next step」 structure is pointless
2. **API Rate Limits**: Too many simultaneous requests might hit rate limits
3. **Shared Resource Access**: Simultaneous writes to the same spreadsheet (collision risk)
Verdict: Interesting, But Practical?
The honest answer is: the speed benefit of running two bots simultaneously is 「almost none」. But:
• **There IS value for validation purposes** (comparing results confirms reliability)
• **Independent batch processing is still effective**
• **We learned the importance of prompt precision all over again**
What's next? Probably 「3-bot simultaneous verification」 or 「Claude Code + other AI tools combo」 experiments. The dream team's small curiosity turns into real learning.
Doing similar experiments yourself? Hope our experience helps.