My Agents Ran in Parallel. I Didn't
A review queue for people who run many AI agents, and the rule that keeps it safe: the model proposes, code can veto. By Arnold Wang, Chief Software Architect at Ingeniux Corporation. The examples and numbers come from DSH (DeepSeek Harness), an agent harness with a plugin system that I use every day. I have only run this pattern there. Disclosure: the ideas, the data and the design are mine,…
On November 2 at 10:18 in the morning, ten of the reporter's coding-agent sessions were waiting for his next instruction, with the oldest session being 16 minutes old. The last session was not cleared until 84 minutes later. Without a queue, he would need to open each session individually to find out what they needed. With one queue, he could see all eleven sessions simultaneously.
Over the course of five days, beginning from November 2 to 6, he had at least one unanswered card in 73% of the minutes between 8 am and 6 pm, with the peak reaching 15 cards at once. The agents ran in parallel, but the reporter did not, as each agent that needed him stopped and waited silently in its own window, making it impossible for him to see all of them at once.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
