Do newer coding models end up training on the AI slop generated by older models?

A new concern emerges that newer coding models may be training on code generated by older models, potentially amplifying errors and 'slop' from both AI and human-written code. The question highlights how attention mechanisms could perpetuate average-quality outputs, raising doubts about the quality of training data for AI coding tools.

As old coding models were writing and pushing code to public, does the new coding models train on them because generated code by old coding models were not so good? How much do we assume human written code for training the original models is free of human made slop? We all know we all take shortcuts and have code we are not proud of. If they are I deed averaging machines, is it possible their output is the average of human output? I'd argue that the way attention works plays into the slop too