Your Agents Are Cheap. Maintainer Attention Isn't.

In January 2026, curl shut down its bug bounty program. Six years, $86,000 in payouts, and a steady stream of real vulnerabilities — ended not because the money ran out, but because the signal did. Historically, more than 15% of submissions turned out to be confirmed vulnerabilities. By late 2025 the rate was closer to one in twenty or one in thirty, and submission volume had spiked to eight times normal. The queue was full of long, confident, completely fabricated reports — one came with GDB sessions and register dumps referencing a function that doesn't exist in curl at all.

Here's the uncomfortable part: the people generating that flood aren't villains. Many of them are engineers like you, running agents like yours, pointed at repositories like the ones your product depends on. AI collapsed the cost of producing a contribution to near zero. It did nothing to the cost of reviewing one. Every economic system with that shape — cheap to emit, expensive to absorb — turns into a spam problem, and the absorbing side is a volunteer who was already unpaid before your agent showed up.

The asymmetry nobody priced in

A pull request has always been a small transfer of work: the contributor does the thinking, the maintainer does the verifying. The system worked because both sides paid a real cost. Writing a patch took hours, so the volume stayed proportional to the number of humans who cared enough to spend them, and the effort itself filtered out most low-conviction submissions.

Agents broke the proportionality. A contributor with a coding agent can open twenty plausible-looking PRs in an afternoon. Each one costs the maintainer the same twenty to forty minutes of review it always did — reading the diff, checking the claims, testing the edge cases, writing feedback. The submitter's cost per contribution dropped a hundredfold; the reviewer's didn't drop at all. Worse, review got harder: AI output looks polished, compiles, and passes lint, so the cheap surface signals maintainers used for triage no longer discriminate. Analysis of AI-generated PRs finds roughly 1.7x more defects than human-written ones — arriving at far higher volume, wearing better formatting.

The results are no longer hypothetical:

• curl ended its HackerOne bug bounty after AI slop drove confirmed-vulnerability rates below 5%.
• Jazzband, the Python project collective, shut down entirely, citing unsustainable volumes of AI-generated PRs and issues.
• Ghostty went from "disclose your AI use" (August 2025) to zero tolerance with permanent bans (January 2026). Mitchell Hashimoto's summary: before AI he saw one bad PR every six months; now it's every other week — and he's careful to note his maintainers use AI daily. The policy is anti-slop, not anti-AI.
• tldraw auto-closes all external pull requests. Gentoo banned AI-generated contributions outright. NetBSD classifies AI code as "tainted" pending approval.
• GitHub is openly weighing a pull request kill switch — letting maintainers disable the core mechanism of open collaboration, because for some projects the mechanism now delivers more noise than value.

Read that list again as an infrastructure engineer. These aren't fringe projects making ideological statements. They're load-bearing pieces of the software supply chain amputating their own intake channels to survive.

Why this lands on your roadmap

It's tempting to file this under "sad for maintainers, not my problem." Your package.json disagrees. The transitive closure of a typical production service touches hundreds of open source projects, and the Tidelift maintainer surveys have shown the same number for years: about 60% of the people maintaining them are entirely unpaid, and 60% have quit or considered quitting, with burnout cited by 44% of those who left.

The slop flood attacks that population at its weakest point — discretionary attention. Seth Larson of the Python Software Foundation put it plainly: wasting volunteer time on work they don't love is the surest way to burn maintainers out. When a maintainer burns out, you don't get an incident alert. You get a dependency that quietly stops receiving security patches, a CVE with no one to triage it, or — the failure mode Daniel Stenberg keeps warning about — a security team so numb from fabricated reports that it misses the real one. The XZ backdoor taught everyone that an exhausted maintainer is a supply-chain vulnerability. AI slop is manufacturing exhausted maintainers at scale.

There's also a nearer-term, selfish reason to care. When projects respond by auto-closing external PRs or banning AI-assisted contributions, your ability to get fixes upstreamed disappears too. The patch your team needs merged into that HTTP library now sits in a queue behind a wall built to keep out people who behave exactly like your agents do by default.

Loading…