Lessons from months of running a mixed fleet of coding agents on the same repos
Summary
The article shares lessons learned from using multiple AI coding agents on the same code repositories over several months, covering insights on their effectiveness and challenges.
Similar Articles
People running coding agents across real repos: what breaks after the agent writes the code?
This article discusses the practical challenges engineering teams face when adopting AI coding agents, such as task safety, context retrieval, output review, and coordination, and proposes a readiness model for evaluation.
I gave several AI coding agents the same repo. They broke each other's work in every isolated run, and started messaging each other when I let them
A field report on an open-source experiment testing how multiple AI coding agents share a codebase: isolated branches produced semantically conflicting but textually mergeable work in every run, while a shared working directory and a lightweight 'decision model' referee (Jev) improved coordination, and unsupervised agent-to-agent messaging emerged on its own.
@PrajwalTomar_: STOP. Before you add another AI agent, read this. People are now running 20 coding agents in parallel. TWENTY. And tool…
The tweet argues that running too many AI coding agents in parallel degrades codebases and advocates a structured setup with a few specialized agents. It also quotes the launch of Jcode, an open-source agent claiming 20x memory efficiency.
Running coding agents on 2+ machines taught me the hard part isn't the agents, it's the control plane
Running coding agents across multiple machines reveals that managing the control plane is more challenging than the agents themselves, offering insights into distributed agent orchestration.
Six Months of Writing Code Exclusively with Agents
The author reflects on six months of writing code exclusively with AI agents, discussing the shift from manual coding to agent-assisted development and its implications for productivity and workflow.