On January 30, 2026, FFmpeg named the PR (#21595) submitted by AMD on the X platform as being suspected of automatic AI generation, which attracted community attention. This PR aims to add AMD HIP SDK support to Windows platforms. FFmpeg maintainer @quink pointed out issues such as inconsistent code style, irrational logic, and the existence of unused constant variables during the review, and believed that the overall performance was "similar to automatically generated code".
Related post link: https://x.com/FFmpeg/status/2016981960015437994
PR link: https://code.ffmpeg.org/FFmpeg/FFmpeg/pulls/21595
Maintainers questioned the quality of the PR and the suspicion of "automatic generation". During the code review of the PR, @quink, one of the maintainers of FFmpeg, made severe criticism of its quality, pointing out that there were many problems, including:
1. The coding style does not comply with the existing specifications of the project
2. The logical structure is unreasonable
3. Unused constant variables (including const of 8) appear. @quink stated in the comments that the overall look and feel of this PR is "similar to automatically generated code", and accordingly questioned its reliability and maintainability. This statement also became one of the focuses of subsequent discussions.
The PR author's response triggered further controversy. In the subsequent discussion, PR author Steven Xiao responded to relevant questions. However, he did not directly deny the possibility that the code was automatically generated by AI, but focused on explaining that some relevant Wiki documents were written by him four years ago, trying to prove his long-term participation and understanding of the relevant technology stack. This response was viewed by some developers as a "red herring" rather than a direct response to the core question of "whether to use AI to generate code."
@quink later pointed out in the discussion that this response did not address specific concerns about code style, implementation quality, and suspected auto-generation traces, further fueling the controversy. In other historical discussions, community members also mentioned:
1. The PR author has been pointed out to have quality issues in multiple previous submissions;
2. Its frequent use of continuous emoji expressions in comments is considered by some developers to be similar to the current common AI-assisted writing style;
3. In another discussion, the PR author once said that he "uses AI assistants in almost all aspects of his work", but emphasized that the final code is reviewed and edited by himself.
The role of AI-assisted programming in open source projects is once again discussed. Although in subsequent exchanges, the two parties have softened their words and attitudes, and technical communication continues to advance, this incident has triggered a broader discussion in the open source community about how to constrain and standardize AI-generated code.