The post says open source contribution has turned into a kind of career currency on GitHub, and AI coding tools now let people mint that currency cheaply by spraying projects with tiny or shallow pull requests. The author’s concrete example was a newcomer opening three separate PRs for spelling and grammar fixes in comments. The changes were correct, but the author closed them anyway because the point was not improving the project. It was accumulating visible contribution credit while shifting review cost onto maintainers.
People mostly landed on a blunt point. A
PR is not free just because the patch is harmless. Someone still has to read it, decide whether it fits the project, think about precedent, and own the merged result forever. That was the throughline across typo fixes, fake perf wins, speculative bug reports, and AI-submitted security issues fishing for CVEs. Several maintainers said the flood is already real, often a few times a week, and that the painful part is not only bad code. It is the entitlement to volunteer attention.
The sharper takeaway was that this is less about AI quality than economics and incentives. Even decent AI output becomes spam when the submitter has near-zero marginal cost and the maintainer carries all the review burden. That is why several people defended rejecting technically correct but trivial fixes. Accepting them teaches contributors that low-effort drive-bys are welcome and turns a repo into an inbox. The comparison to
Hacktoberfest came up often. Once contributions become legible status, they get Goodharted.
A second thread pushed the argument beyond repository hygiene into hiring. Multiple commenters said open source contributions are no longer a reliable hiring signal, especially when candidates can cheaply generate visible activity. Some went further and said performative
OSS work may now be mildly negative signal, because it shows optimization for the metric rather than for useful work. The practical response was not “ban AI.” It was “make contributors own what they submit.” A few maintainers shared policies that allow AI-assisted PRs if the person can explain the change, follows issue templates, and is not obviously spraying random repos.
There was still discomfort with turning the defense itself into more AI. People liked automated filtering and auto-closing for template violations, but disliked the idea of contributors being forced into conversations with patronizing review bots just to earn human attention. The line many settled on was simple.
CI and static checks are fine.
LLM gatekeepers that simulate human review are a worse experience and risk punishing sincere contributors along with the slop merchants.
neilalexander.dev