Comment by HappMacDonald

I find that shutting off one's brain and just slogging through a task leads to work quality on a par with AI slop.

So if one really is as uninterested in the quality of the output as you suggest, perhaps it might actually be better to dump the problem into Claude/Gemini/Cleverbot and just copy/paste/act upon the results verbatim and then mark the checkbox as "done" and move on.

For me personally, the pain of such efforts is ordinarily from making sure that the output is correct when the input is largely guesswork or speculation that always leads to hunting through a morass of poor documentation of some library or seeking a workaround to some irritating problem or rolling the dice on what the risk to various decisions might prove to be over the future: "eh, duct tape this and it ought to hold".

And most notably that doing more of this work correlates to an exponential rise in the volume of similar work that will be required down the road to maintain the same results.

Those are often exactly the time one would be best served by taking a step back and questioning the entire framework that supports the busywork in question. Perhaps starting from scratch or making some huge change would reduce the garbage portions of the effort and keep them from further proliferating?