From a lot of posts here I get that working as a dev in the US is now a total shitshow. But to give you my European perspective (I work in Germany): AI adoption hasn't been as rapid here. People and companies are more skeptical about it, compared to the US.

I work as a web developer for an employer that is cautious about AI. I can use it, but I am not forced to. Tried it excessively for a few months (agents writing my code, playing a glorified manager and all of that jazz), but I noticed my own skills atrophying and me losing the general grasp of what my code actually does. And even though everyone and their dog claim that the models get better with each new release, I still run into hallucinations way too often. If you are very experienced in a field and you've been doing it for over a decade, you notice all of the small inconsistencies and bullshit answers - much quicker than a junior dev who didn't have that experience yet.

So nowadays I only use Gemini for tool and library research or really simple boilerplate code. For everything else my own brain is the better solution. I am not actively against AI as a technology, but extremely opposed to paying a subscription to some techbro billionaire's company to keep doing my job. Fuck Altman, Elon, Jensen and the Zuck.

If Ed Zitron is correct in all of his calculations, the frontier models will get so expensive they'll become unprofitable for a lot of companies, so it would be a stupid decision to rely on them. I am looking forward to one day host good models on my own machine - though that day is not today, when capable GPU's still cost thousands of dollars.