> I wonder how much user faith in ChatGPT is based on examples in which the errors are not apparent ... to a certain kind of user. I think this is it, basicall

> I wonder how much user faith in ChatGPT is based on examples in which the errors are not apparent ... to a certain kind of user. I think this is it, basically. A certain segment of programmers are simply not skeptical enough. The same sort of skepticism that will keep jepsen.io going until the heat death of the universe debugging faulty distributed consensus algos. My coworker recently generated a bash script with it. The shell was terrible, full of novice shell anti-patterns. But this makes "sense" — what was it trained on, after all? Most shell in its training set would be of low quality, because most shell is of low quality. The script was "determine lines added/removed by author in git repo". The `git log` call it made was bizarre — emitting way more information than it needed to — which it then subsequently filtered out extraneous bits with `grep`. … but it could have just never outputted them in the first place. Things like `| grep | awk`, lack of quoting where needed, etc. Multiple loops through the log output where one would do (and having to slurp it into memory because of that, where a simple git log | <process output> would suffice). The regexes to match the log output were wrong, resulting in the wrong output — but that didn't stop multiple people from going "nice" or "cool script" first.