In many cases, the final product required few, if any, human changes.
This is the crux tho. They try to cleverly hide it, but this wording basically translates to "it constantly makes mistakes, but sometimes its half decent".
In reality people are not checking the code at all or just passing it to another LLM that is supposed to find bugs. Sure there are people who use it wisely and im sure they will immediately cry "but it works for me" but lets be real, this will be used by people who do not understand what they are doing at all and are completely incapable of even understanding the code that the thing writes.
And they will push this code into production environments in government, healthcare, military and infrastructure deployments.