"I don't know if my job will still exist in ten years" by Sean Goedecke https://www.seangoedecke.com/will-my-job-still-exist/
Basically took the words out of my mouth. I wish I had written this.
0 Boosts
0 Likes
@nolan One thing this post just blindly assumes is that AI costs stay as affordable as the are today. There are many industries today where human labor could be automated away, but doing so would be more expensive than having humans do it (sometimes abroad). I'm not saying that's an optimistic option to look out for, just that the author didn't consider this argument in the post.
@tomayac Correct, at this point I think it's the only deus ex machina that could save the profession as we've known it. I don't think the odds are zero, but they do seem low given the effort thrown at open-weight models, GPUs, etc. In any case yeah he doesn't acknowledge it.
I've always felt at some level that the job of a software engineer (a good one, anyway) is to automate yourself out of a job. If you're running some step manually every day, write a Bash script. Set up alarms. Reduce the bus factor. Etc etc.
I guess it never occurred to me that we might reach some plateau where nearly everything could be automated, but it feels like we're reaching that endgame now. I recognized this about a year ago and it made me extremely depressed.
I would love to say that this is just AI psychosis, but at this point I'm asking Claude to write the code, review it, monitor the deployment, identify structural flaws, design the next architecture, etc. My taste and judgment are there, but how much longer can I cling to that? And if I'm saying that, how must *juniors* feel?
Even a year ago you needed a lot of elbow grease, but the mistakes and hallucinations have really gone down since then. The trajectory seems terrifyingly clear.
> Even a year ago you needed a lot of elbow grease, but the mistakes and hallucinations have really gone down since then.
Do you ever wonder how much of that is because of *us*, as in (this is not a dig, I swear, I respect you Nolan, one of the very, very few people who, even if begrudgingly, embrace AI) people like you who keep letting Anthropic distill them into evolving their models by working with it, reviewing etc.
This is something that would, quite literally, keep me up at night if I was forced to work with Claude at work (which, to be fair, transitively I still have to, because you can't avoid others' uses of the AI when you work on a shared thing).
@flaki It's a fair point. To some degree, undoubtedly yes: I have helped bring about the very situation that might put me out of a job. But also: I already did that by writing a blog and putting OSS code on GitHub. And AIUI the big gains in coding agents lately are more due to RL (i.e. training against benchmarks). Plus if I hadn't done it, many others would have taken my place. I don't regret my actions because ultimately I don't think it's reasonable to blame myself for this situation.
@nolan my view: thought is essential. i think by coding. i normally code at the speed i think. when i find myself thinking faster than i can code i recognize this as a flaw in the code structure. so i fix that, then back to coding at the speed i think.
not much space for a chatbot in that process. maybe outside it, identifying problems.
i also think that the people whose careers are in danger are those who allow the chatbot to get in the way between their thoughts and their code.
@nolan I can help: this is AI psychosis
@nolan Hmm, I've been telling my family this too, but my horizon is more 5 years.