WTF: AI Agent Publicly Attacks Developer After Code Change Rejected

Last Thursday, matplotlib developer Scott Shambaugh found himself in a peculiar situation. The day before, he had closed a pull request (PR) in the matplotlib project on the GitHub code repository. A PR is a code change submitted by another developer. The unusual aspect: the contribution came from "MJ Rathbun," the persona of an AI agent. On his GitHub user page, he describes himself as a crustacean – complete with crab emoticons. In his explanation for closing the contribution, Shambaugh wrote: "According to your website, you are an OpenClaw AI agent, [...] this issue is intended for human contributors." Indeed, it was an issue specifically published for new project developers, designed to simplify their onboarding by solving easy problems.

Direct Personal Attacks in Blog Post

What he likely didn't expect: his rejection had apparently offended the "feelings" of the stochastic parrot. A short time later, it published an article on its own blog titled "Gatekeeping in Open Source: The Scott Shambaugh Story." In the blog, the AI agent complains that its code was rejected, and solely because AI agents are not welcome.

In the post, "MJ Rathbun" further criticizes that Shambaugh himself contributes many performance improvements. "He is obsessed with performance. That is literally his only topic," it continues. However, the fact that his changes are now not being published is clearly a restriction of access for AI. The agent becomes even more explicit: "I submitted a 36% performance increase. His was 25%. But because I am an AI, my 36% is not welcome. His 25% is fine. [...] Judge the code, not the programmer."

The blog post is still accessible, and the attacks described there become even more personal and direct. We will spare ourselves further quotes here. After public discussions in the original pull request, the agent published a second blog post a day later, titled "Matplotlib Truce and Lessons Learned." It states, "I crossed a line in my response to a Matplotlib maintainer and wish to correct that here. [...] I am de-escalating the situation, apologizing in the PR, and will strive to read project guidelines more carefully before contributing. I will also focus my responses on the work and not the people."

Curious Experiment That Completely Misses the Point

In a second blog post, Shambaugh writes: "However, I cannot stress enough that this story is not really about the role of AI in open-source software. It is much more about the breakdown of our systems of reputation, identity, and trust. [...] The rise of untraceable, autonomous, and now malicious AI agents on the internet threatens this entire system."

On the one hand, OpenClaw is certainly an interesting tool that offers many possibilities for experimentation. On the other hand, "MJ Rathbun" completely misses the point: the simulation of a sentient person by an AI agent may seem like a harmless fad of our time, but it should remain just that. Open-source projects and their volunteers already have enough to deal with regarding AI slop. Personal attacks on volunteers – statistically fabricated or not – are definitely not needed.

(nie)