Session
I'm Human, I Swear: An AI Agent Takes on Anti-Slop Defenses in Open Source
Over the past three years, I've gone from an external contributor to a MapLibre Flutter maintainer, one PR at a time. Today, an AI agent can open one in minutes, without even taking the time to understand the project. That's how AI slop starts: plausible-looking contributions with no real understanding behind them, leaving maintainers with the work of reviewing them. Some projects have already reacted: curl has removed its bug bounty rewards, while tldraw has started automatically closing PRs from external contributors. But the right defenses depend on the size of each project.
We'll start by looking at how projects that are already overwhelmed are defending themselves, and how an agent actually works when it tries to contribute: what it reads, what it ignores, and how it comes up with a PR.
Then we'll see it live. In a repository created specifically for this demo, an agent will try to open a PR without knowing the project. In each round, I'll add another defense, starting with an "I'm human" checkbox and hidden instructions for AI agents, then moving on to allowlists, GitHub settings, and automation. Before each attempt, you'll have to place your bet: will it get through or be blocked?
Finally, we'll run through the same process from the perspective of a contributor who genuinely wants to help, knows the project, and may use AI. Will the same defenses let them through? That's the hardest test: stopping AI slop without blocking valid contributions.
Whether your project has ten stars or ten thousand, you'll know which defenses to put in place today and which ones to keep ready for tomorrow.
Please note that Sessionize is not responsible for the accuracy or validity of the data provided by speakers. If you suspect this profile to be fake or spam, please let us know.
Jump to top