AI slop and the Aussie devs fighting to maintain open source

Software development is in the middle of its most exciting shift in a generation. Generative AI is transforming what developers can build: taking repetitive, mechanical work off their plates and freeing them to focus on the design decisions and higher-value problems only humans can solve.

AI is also lowering the barriers to entry for software development. The number of Australians building software on GitHub increased 24% last year and now sits at 2.2 million developers. Globally, more than 80% of new developers on GitHub are using AI-assisted tools, such as Copilot, within their first week. That’s a good story for software development, and particularly good for open source, which has always thrived when the barriers to contributing get lower. Open source projects are the invisible foundation that helps you bank online or on your phone, shop online and access government services — and the more developers contributing to them, the better.

But every wave of democratisation brings growing pains. In the mid-1990s, when dial-up ISPs sent waves of newcomers onto Usenet, unfamiliar with the norms, the community called it the ‘Eternal September’ — the moment the newcomer flow stopped being seasonal and became permanent.

Open source is having its own Eternal September right now. AI has collapsed the time it takes to generate code, but the time it takes to review it hasn’t moved. For ‘maintainers’ — the volunteers who manage open source projects and act as gatekeepers for growth — the pull request pipeline is flooding over. The maintainer community has a well-known moniker for many of these contributions: ‘AI slop’, or plausible-looking, sometimes-correct contributions created by well-meaning amateurs, but sent at high speed and high volume.

Australia happens to be home for maintainers working on some of the world’s biggest open source projects. They’ve been at the forefront of this AI-generated wave but are also building the solutions to keep open source projects viable and thriving.

Lucy Liu, numerical and scientific computing maintainer in Melbourne, said AI has “flipped the script” on her work. Where once the challenge was helping newcomers write good code, now it’s filtering the sheer volume of plausible-looking code arriving at the door.

Michael Neale, a Block engineer who helps maintain an open source AI agent used by more than 100,000 people worldwide, says he is “fighting fire with fire” by using AI to help triage the AI.

That’s the pattern I keep seeing everywhere in open source right now: when the community hits a wall, people build ladders. Maintainers aren’t shrinking their role in response to AI, they are still the ultimate experts who decide what the AI got right and what it got wrong. The judgement calls — what gets merged, which direction a project takes, what ‘good’ looks like for the project — still require a person.

Despite the individual ingenuity of maintainers, there’s more we can do as a community to support them. If maintainers burn out or walk away, projects will falter and this will undermine the collective work of thousands of contributors. Maintainers, including those based in Australia, have been instrumental in GitHub’s rollout of new tools to help them stem the flow: repo-level controls that let maintainers restrict pull request creation to collaborators, or disable pull requests entirely; pinned comments on issues; banners that reduce ‘+1’ comment noise; temporary interaction limits; and faster pull request diffs and issue navigation.

None of these tools by themselves will solve the issue, but they’re part of an effort to ensure maintainers are seen and heard. Australian developers have been at the forefront of this.

The AI revolution can’t be undone and, taken as whole, has been a huge benefit to software engineers. As another Aussie developer, Melbourne’s Harlan Wilton, told us, the use of AI agents has freed him up time to improve his understanding of high-level architectural decisions. “I’m learning a lot more and taking in a lot more …understanding conceptually all the programming paradigms much more deeply than I would otherwise,” he said. By rolling out AI agents to support human engineers, we can support the critical role that open source maintainers play in building what the world runs on. But we can also keep making software development even more of a collaborative, experimental, engineering buzz — the reason why us developers loved doing it in the first place. *Ashley Wolf is Director of Open Source Programs at GitHub.

Top image credit: iStock.com/Aree Sarak