It's an unspoken rule that large pull requests are poor etiquette. Traditionally, Agile teams break features into manageable slices to make it easier for developers to tackle the logic and for peers to actually review. That is until the adoption of agents. Now developers are shipping end-to-end features in a fraction of the time, but the result is often an unmanageable wall of code that is discouraging to even open. These "Agent-PRs" almost guarantee a reviewer's eyes will glaze over, leading to a quick "LGTM" that misses subtle logical flaws. This tension has become so high that some open source maintainers have moved to outright banning agent-authored contributions.
Before the invention of pull requests, contributors would email maintainers their code changes or ask maintainers to pull updates from the contributor's repository. In 2008, GitHub introduced pull requests, giving teams a structured way to propose, discuss, and review changes before merging them. For almost two decades, the pull request provided the perfect workflow. But now the very structure that enabled collaboration is cracking under the weight of AI-generated velocity.
Prominent open source developers, such as Rémi Verschelde and Jeff Geerling, have taken to social media to express their concerns.
But their sentiment isn’t isolated. Many developers feel this way, and faced with deluge, many have decided to take drastic measures to protect their codebase and sanity by banning AI-assisted contributions completely.
However, as more developers are encouraged to use AI throughout the software development life cycle, running away from AI-assisted pull requests only hurts open source by closing the door to potentially valuable contributors. Angie Jones, my former manager, argues that closing the door isn't the solution. Instead, she advocates for better guidelines for both agents and humans.
She's right. Closing the door isn't the answer, and clearer guidelines are a step forward.
But even with better guidelines, the review process remains fundamentally unchanged. Maintainers are still staring down massive PRs, trying to make sense of what an agent produced.
To remedy this problem, GitHub recently released a feature giving maintainers the power to set contribution limits, starting with a PR cap for outside contributors and an allowlist for trusted ones.
Many maintainers are excited about this release, but a PR cap is a short-term band-aid. I think it’s a good move, but our industry needs a complete overhaul of how we handle code reviews.
As a maintainer of goose, an open source AI agent, I have spent 2025 and 2026 sifting through pull requests. I do not want to limit who can contribute. I want everyone to be able to contribute. The goal is to understand the context of a PR at a fast pace, even if an external contributor uses an agent to build it. Open source needs infrastructure built to support both human-authored and agent-authored work.
I joined Entire because our convictions aligned: we need to fix the structural breakdown in open source. Right now, we have built a CLI that provides a system of record that captures the context behind agent-assisted code changes. The record gets stored as a This serves as the baseline for a new era of developer tooling that can help our industry move towards:
- Shifting from code review to intent review: This means instead of parsing 500 lines of syntax, reviewers will start with intent by examining the prompt, the session transcript, and the reasoning behind key decisions. This allows reviewers to focus on the problem being solved and whether the right calls were made along the way.
- The ability to search for the why: Developers and maintainers can ask why a change was made a certain way and receive an answer derived directly from the agent session context.
- Infrastructure for the AI-native velocity: The open source community needs a foundation that can support the massive volume of human and agent contributions without falling under the weight of the traffic or causing platform outages.
To explore the full scope of what we are building, you can read about our vision and learn more about where we are headed.
Open source is where innovation happens. It is how large companies thrive, relying on thousands of open source dependencies. Yet if we shut down external contributions out of fear or fatigue, the community disappears, leaving the ecosystem at risk of stagnation. Developers are genuinely excited to contribute because agents have given them the confidence to do so. We need to embrace that momentum by building an infrastructure designed for the AI-native lifecycle.