I came across this article on Medium by Marina Wyss, a data scientist at Twitch.

I still find it hard to determine what's real when it comes to programming using AI tools. Social media is full of people declaring that they don't write code anymore, that they trust whichever agent they're using to do the job. It could be I'm just using it wrong.

It's critical to remember that someone is responsible for the output. This part gets left out a lot in social media posts venerating the abilities of gen AI tools. Marina writes:

Because here’s what doesn’t change regardless of how good AI gets: when something breaks in production — when there’s a security breach, a compliance violation, or an outage that costs the company boatloads of money — someone is accountable.

AI doesn’t get paged at 3am. You do.

AI doesn’t get called into the incident review. You do.

AI doesn’t explain to leadership why customer data was exposed. You do.

Don't forget it.