Esa Kylli

long write up, hope it helps: one thing you learn as you get older is that judgment and responsibility are extremely valuable in the marketplatce. in short it's the difference between "being able to build something" and "being able to figure out what is the right thing to build in the first place", and being accountable for it. the reason LLMs can't ultimately replace humans is that they have no reliable judgment and can't be held accountable. if you go to an AI agent and say "build me a renderer", it will produce *something*. Being able to tell whether it's the right thing (and even specifying what the right thing is in the first place), requires domain knowledge. For software, these are the fundamentals of how computers work (OS, memory, CPU, GPU, networks, etc), as well as a lof of expertise in managinc complexity. so if I were 20 again and was just starting my career, I would still learn to program by hand, and focus a lot of effort in learning the hard fundamentals of my field, by building hard projects (an OS, a compiler, a renderer, a physics engine, whatever interests you). I would use AI as a teacher instead of a tool to do things for me. hope it helps!