Four coding agents simultaneously appears quite productive. But who is monitoring all their output?

In this week’s Leaderboard newsletter about Halloween, there was a feature where you can launch four coding agents side by side and communicate with any of the four as necessary. I understand why this is attractive, but this weekend, I tried something similar with three tabs of terminal window and three different tasks related to the budgeting application that I am building, and the bottleneck immediately changed. My agents finished their task much faster. I became the bottleneck in this case, because analyzing three diffs carefully took me more time than analyzing one, and my concentration was thinning out by the minute. It was a fortunate accident when I found a duplicate helper method.

It turned out to be much better when each agent was restricted to a specific area in the codebase, and the size of the tasks was such that it could be analyzed in five minutes. Conflicts started when two agents were working in the same file.

The basic rule that I follow at present is never to generate more agents than I am able to check prior to lunchtime. Quick generation will not matter much in case there is no quick checking.

How many parallel agents do you actually generate? Do you have a review process, such as checking tests first, to maintain high quality amid large numbers of diffs?