I used to be very pessimistic about using LLMs in my work. I did it now for the last couple of weeks and it’s really astonishing what you can achieve with it nowadays.
And when I read about “code is cheap now” recently, I immediately knew what it meant. The code for a fairly good working feature is just a good prompt away with something like Claude Code. And it’s up to you to review and test it. Things you should have been done before anyways, but the coding part is much more compressed, coming from a 10+ hour task down to maybe 30 to 60 minutes. That’s huge.
On the way, I was confronted with a little existential crisis … What am I here for? Couldn’t anyone do the same? Technically, yes, as the barrier is much lower now. So it comes down to my own knowledge and being able to have meta skills like having a sense for good software architecture (compared to working code), critical thinking, being able to find the right solution for the problem and similar. While LLMs mimic these skills, they actually lack them – at least for now.
What’s important to not staying still, but instead learn new things and improve the own skillset, especially in areas LLMs are not good (yet). Chris Gregori explained it very well.
I’m just her to tell you: if you are anxious about LLMs in connection to your job and thus dislike everything it produces, that’s okay. Just keep it a chance, look for personal benefits in using it and create a list for yourself to note things you can still do better. And then even further improve these things. LLMs are probably here to stay, even if the bubble will burst spectacularly, but so are you if you focus in your skills and experience.
Leave a Reply