AI sucks in my domain. One of our competitors uses it and they say they get amazing results when senior people use the tools but can’t give them to juniors because they keep messing up.
I’m glad it doesn’t work in my niche to be honest. I’m in the frontlines debugging broken code and the last thing I need is bloat. This would actually slow me down a lot. I find it pretty shitty at diagnosing even small pieces of code, and I can’t try stuff like Claude Code because I’m not allowed to transfer some of the code we use to the cloud because it’s under NDAs. But if it can’t get the simple stuff right, I can’t trust it with the keys to our Lambo.
I stay informed about it all to know when/if I should quit software engineering and do something else, but it seems fine so far. It looks like I won’t be able to take it easy in the future and go back to pissing webapp code, which means less opportunities, but oh well.
Oh and I do know two companies that mandate LLM usage. Somebody from my current company left for one of those and hates it.
Or AI usage might become too expensive as energy costs rise, datacenter equipment becomes more expensive, monopolies emerge, investor funding dries up.
Putting all ethical concerns aside, the bad code written by ClaudeGPT could also finally cause a software quality decrease and explode in the faces of the companies using them, making any velocity gains dubious. Services may also start enshittifying.
There’s also the legal aspect which is not fully settled. The dependence on US companies which might not remain OK forever in the rest of the world.
It is to be noted that, while programmers and executives are claiming they see gains, the science is not settled. Studies so far seem to indicate the contrary is observed in practice, although it remains to be seen if it stays that way in the future because, apparently, results got way better with Claude Whatever 4.6.
I’m not predicting a collapse, just saying these are plausible scenarios. And if any of them comes true, even if there’s no collapse, there will be a spot for actual software engineers that refused to use it all along and remained sharp, and they won’t be “left in the dust”.
It sure is a bad time to get into webdev though as a LLM-skeptic. You really have to choose your jobs to avoid writing code millions of developers wrote before, because there’s a good chance you’re working on something nobody needs and/or that slop machines can do it faster than you.