The analysis covers the SuiteCRM 7 repository and examines more than 1 million lines of code across 5,500+ source files .
For context, the official SuiteCRM repository is a large, mature open-source project with a long development history, so automated AI-authorship detection is particularly interesting—and potentially difficult—to interpret accurately.
Are there particular findings in the report that deserve a closer manual review, or do you think the methodology produces too many false positives for legacy/minified/generated code?
It’s interesting to see - very little action items included and business wise probably of very low value.
Context wise, the front page already explains a bit more:
AI-generated code is not inherently a problem — it is a powerful tool that accelerates development. https://slopcodemonitor.ai/
As you wrote, judging by the methods, there could be quite a few false positives.
Probably easy to find out, by just downloading a ca. 2 year old version of SuiteCRM and run the check there.
Overall, when the CRM runs smoothly, the business processes and requirements are supported and automated, is AI slop worse than poorly written code from a human with little dev experience? And is handwritten code better than AI code generated by an senior software architect with deep understanding of what she/he is prompting?
In the last few months I’ve seen so many ‘AI slop’ projects on Github.
Several vibe coded customer portals and CRMs generated by non technical people without any knowledge of software architecture, software security, CI/CD.
And a few good examples where experienced dev teams for the first time in company history outpaced the designers and product owners, since they are much more productive with AI now.
Eventually, LLMs have been trained on human written code. There have always been good and bad examples of human written code.
The whole topic should be regarded more nuanced.
