I think the skills I’ve been learning over the last 13 odd years will be valuable in this new world. However, I’m not sure there will be many people like me needed, or whether I’m good enough to be needed at all. I think I need a fundamental re-skilling. More on the artistic side, more focused on talking with people and gauging what they want, then translating that into a prompt an AI can work with. Then let it simmer, stir the pot every few minutes, click some “yes” buttons, make a few suggestions, and just ship it.
I think the kind of productivity we as a society could gain from that is something we might not be prepared for, especially in Germany, where everything takes an insane amount of time. I also think that besides new regulations, I’m not even sure we have the bandwidth to come up with as many ideas as we could write code for. So I struggle to believe that we will be paid as much as before. I rather believe that human jobs and craftsmanship will start to rise again. When everybody can create AI slop and the barrier to entry is reduced so much, we won’t be satisfied with artsy, AI-generated stuff.
I also never understood the argument about code maintenance and organization with LLM-generated code. At some point, I won’t even look at the code, and if there is a bug, the LLM should figure it out. So entire architecture and code-style discussions might just become irrelevant. I know this is very outcome-focused.
I think we, as developers, need to be careful. I know everybody and their mother is writing and talking about AI — in this case, literally — which is mind-blowing in its own right, and now I’m jumping on the bandwagon too. However, I’m not doing it for the hype. I truly think it changes the game substantially, and I don’t think people are seeing this clearly enough.
The message I’m trying to convey is this: AI will change the way we work with code at every level of the stack. If you still think it might be a fad that will blow over, let me say it again: it’s not a question of if, it’s a question of when — and that “when” will be sooner rather than later.
The thing is, we don’t need superintelligence to automate ourselves away — especially in the business of CRUD-type apps, where the business logic fits on a napkin. The LLMs we have now are already more than capable of doing that. What’s missing is tooling that needs to play catch-up. While it’s still early days, there’s no way it won’t.
It’s like bootstrapping a machine that can build many other machines to make itself faster. That’s the situation we’re in with tooling — which is why the speed of all this is insane. It’s worse than JavaScript frameworks in the 2010s. Back then, a new frontend framework or tool emerged every week. With LLMs, it’s every day. The pace is insane. It’s as if Eliud Kipchoge were trying to run a sub-two-hour marathon, and while running, his shoes were constantly being fitted, refitted, and replaced with more efficient materials.
As it stands now, we don’t need superintelligence to see that most developers will not be developing software the same way they have for the past 40 years. It might be the case that organizational inertia in established companies will take some time to catch up — except perhaps for really large companies. But even they will get eaten if they don’t.