Don't let the machines do it

There is a widespread problem with the stuff people are allowing LLMs to do for them. Specifically, they’re allowing machine-assisted coding in areas where the person writing the prompt couldn’t do the code review. That’s not great. It’s not great because we’re allowing the slop past points where they would have otherwise faced scrutiny.

I’m not going to avoid the uncomfortable fact that I myself have allowed some of the results coming out of LLMs to blow past me. Most of the time I was in a hurry. We should be careful to think about how we use these things. Instead of asking it to make something wildly above your head, you need to be using it to augment what you already know.