Yes, quite extensively. It was quite good. I'm very enthusiastic about it. I just want some data other than my n=1 experience.

The thing about it is the distribution of wins is going to be different depending on the prompter, and that will make all the difference. As experienced devs, we can prompt the AI, and take the wins when they come, but deftly avert the nonsense when we see it, because we know what to look for. For experienced devs like you and I, this is a huge win and will result in productivity gains.

I worry for the junior dev. They can take the wins and feel good, but they’re going to fall for the nonsense every time, because they don’t k ow how to spot it. Junior devs are fed lies and they take them to be truths. Hopefully they are dispelled immediately when they are tested in code, but other lies will persist and be sold as truths. I am worried this is a bigger problem than people might think, cause it’s a feedback loop that can lead to a vicious cycle. We see what vicious cycles have done to social media, and social spheres, imagine what one will do to software systems.

Next year we’ll be graduating students who started college during the pandemic. I would say their programming skills are quite atrophied compared to previous generations. Now I’m worried in four years we’ll be graduating students who started high school in covid times and started colleges using AI chat bots. Who knows if these graduates will even be able to do anything close to what students even 2 years ago could do. They’ll all just blankly stare and reach for their iPhones like they do today, but exponentially worse because they won’t even be able to formulate the search prompt. They’ll need AI for that too.