Are we learning to code, or learning to be replaced on a schedule?

I study naval architecture during the day, and I teach myself programming at night, and some evenings the two halves of my life argue with each other about whether the second one is pointless.

This is not a post with an answer. It's me laying out my actual hedge, and asking what yours is.

My situation, without the drama

I'm a beginner. Nobody is paying me to code yet. The work I hope to do one day is exactly the work AI tools are getting scarily good at: small scripts, simple websites, the junior-level tasks.

Senior developers argue about this from experience. I have to argue about it from the starting line, which is a different problem. They built their judgment before the tools arrived. I'm trying to build judgment with the tools sitting right there, offering to skip every hard part.

My hedge, such as it is

Learn the parts you can't hand over blindly. When AI writes my code and it breaks, either I can read it well enough to find the problem, or I'm just a person copy-pasting errors back and forth. Fundamentals are not nostalgia. They're the difference between using the tool and being used by it.

Build real, finished, small things. Not impressive things. Finished ones. A program that does one job for real people, even if the real people are my housemates splitting a food bill.

Practise checking AI output like it owes me money. I read generated code line by line now, the way you'd count change from a shopkeeper you don't fully trust. Sometimes it's right. The skill is noticing the times it isn't.

The plans I hear from other learners (said fairly, I hope)

Some learners I talk to are going all-in on AI tools. Their argument: the job is changing, so learn the new job, not the old one. Prompting, reviewing, stitching tools together. I think they're half right, and the half that worries me is that review skill has to come from somewhere.

Others are doubling down on pure fundamentals, no AI at all, like it's 2010 forever. I get the instinct. I also think refusing the tool entirely is its own risk, because the people you're competing with won't refuse it.

A third group is quietly quitting, and I understand them most of all. Nobody promised us the ground would hold still. But quitting because the field changed feels like leaving a boat because the sea moved, and I say that as someone literally studying boats.

The fear, said plainly

Here's the version of the worry I don't usually type out.

I am spending my evenings, which are not unlimited, on a bet. The bet is that learning this slowly and properly will still be worth something in five years. Some nights I watch a demo of an AI building a whole app from one sentence, and my little expense splitter feels like learning to churn butter the year the supermarket opened.

Then I remember the demo always ends before the app breaks. Somebody has to be there for the breaking. My whole plan, if I'm honest, is to be employable at the moment things go wrong, because things going wrong is the one part of software that has never once been automated away.

What I genuinely don't know

I don't know if my hedge works. Nobody my age does. We're the first batch of beginners who have to learn while the definition of the job moves, and anyone who tells you they've solved it is selling something.

What I know is smaller: the fundamentals make the AI useful instead of dangerous, and finished projects are proof that survives any argument about the future. That's enough for me to keep going tonight.

So, learners specifically. Not the seniors, not the gurus. Us:

What's your actual plan while the job changes shape?

What are you learning that you think AI can't take, and why that?

If you had to bet one year of evenings on a single skill, which one gets your year?