Linux Foundation's 25% AI Slop Estimate Sparks Firestorm Over Validity and Impact
The Linux Foundation released a report estimating that 25% of new pull requests are low-quality AI-generated code, sparking intense debate. The figure, based on maintainer opinions rather than measured data, has been called into question by critics who argue the methodology lacks transparency.
Thorry claims AI 'ships shit out faster' and increases workload for maintainers, while atzanteol argues AI 'accelerates OSS development in more ways than simply writing code faster'. Mushroommunk cites MIT studies showing productivity remains equal despite overhead from managing LLMs. Currycourier dismisses the study as a flawed survey, noting people are bad at estimating their own speed. Alottafagina highlights the lack of transparency in how the 'AI slop' question was framed, casting doubt on the 25% figure.
The community is deeply divided. While some see AI as a productivity boost, others warn of lower code quality and increased maintainer burden. The consensus is that the 25% estimate is an opinion-based guess, not a measured fact, and that the debate over AI's role in open source is far from settled.
Key Points
#1The 25% AI slop estimate is based on maintainer opinions, not measured data
Thorry and others argue the figure is an opinion-based guess, not a measured count, and that AI-generated contributions often require significant human review.
#2AI boosts productivity but risks lower code quality
Thorry claims AI 'ships shit out faster' and increases workload, while atzanteol argues AI 'accelerates OSS development in more ways than simply writing code faster'.
#3Survey methodology is questioned for lack of transparency
Alottafagina highlights that the survey lacks transparency in how the 'AI slop' question was framed, making the 25% figure questionable.
#4Productivity gains may be offset by overhead
Mushroommunk cites MIT studies showing productivity remains equal despite overhead from managing LLMs and reviewing output.
#5AI-generated code may not be worse than human-written code
CameronDev argues anecdotally, human-written code can also be garbage, so AI isn't necessarily worse.
#6Need for policies emphasizing human oversight
Stsquad calls for policies that allow narrow AI use but emphasize human understanding and ownership as mandatory components of contribution.
Source Discussions (4)
This report was synthesized from the following Lemmy discussions, ranked by community score.