Lately, I have been getting the same question from relatives and juniors which keeps landing in my inbox, and given AI, should anyone still become a programmer, and my answer has settled into yes, and…

I have never taught a CS class, but I started on rentacoder in 2003-04 doing WordPress and Joomla glue where projects started at $5 and stretched to a few thousand, and that bottom-of-the-market grind taught me what the job actually is, which is problem-solving with computers and learning to control complexity while solving those problems, and I have a hard time imagining a future where that skill is worth less than it is today.

Yes, because you have to write the code, and the thing about AI code generation is that it works well enough to rob a junior of the trenches, because the visceral understanding of code comes from being down there writing loops that break at 2am, and if that step gets skipped the ability to read code never forms, and reading is what will matter more in an AI-heavy future.

In plain English; generated systems grow past what their author can hold in their head, which leaves the author conducting an orchestra of sorcerer apprentices that keep fetching water after the order stopped making sense.

So here goes; write the code by hand while learning, and let AI generate it only after you can already read what it made and smell where it cheats.

That reading ability is also why the assembly analogy never sat right with me, and compilers are deterministic in a way models are not, because a for loop lowers to roughly the same machine code for a given architecture every single time, while the same prompt invents a new architecture of excuses every single time, and high level languages removed accidental complexity while generated code often adds it back through wrong approaches and shortcuts nobody asked for. Thats making sure your review queue hangs at least once every few prompts.

And raw typing will matter relatively less, which bums me out a little, because I like the act of coding with metaphorical bare hands, and there is art in naming and shaping a function well, and that pleasure stays even as its market price softens, and other muscles have to grow alongside it.

Communication grows first, since unclear thinking produces unclear prompts which produce unclear systems, and the programmers I know who already read books and write posts adapt fastest because they can say what the system should do in words humans and models both accept. The second muscle is business understanding, because every company over a hundred people has problems which need solving with computers even when they call the team something else, and a programmer who learns how the money moves can keep working as a programmer while picking which real-world problem deserves the next build.

Wouldn’t it be nice if juniors were allowed to stay slow on purpose for a year while everyone else vibes at speed?

The third muscle is architecting, which is a word I distrust after watching architecture astronauts inflict pain, and yet someone has to organise large systems and hold complexity down across years, and that intuition historically came from building small parts badly and then better, which is exactly why juniors must keep writing code and why companies must let them, since bad architects are usually people who never carried a small system long enough to feel it rot.

Using models well splits by experience from what I have seen in my own backlog, and seniors who already know what good looks like do best when they point models at analysis and organising thoughts and small throwaway demos and the bits nobody enjoys like regex and CSS, while keeping API design firmly in human hands, and juniors do best when they fight the temptation to vibe through tickets, accept they will look slower than peers for a while, and use the model to clear accidental complexity like build setups and API shapes so the road to senior gets shorter instead of hollow.

The job market part deserves plain talk, and online job sites are mostly a lottery for juniors which is still worth playing because tickets are free, and personal connections convert far better, so family and friends and family of friends beat cold applications, and the trick is that the family need not work at a big tech company at all, since a retail or logistics or hospital outfit with a hundred-plus staff already has a shadow dev group waiting for someone who can program on top of analyst work.

I started remotely over two decades ago when that path looked shaky, and the market has boomed and busted cyclically ever since, and this bust feels longer because models ate the bottom rung, and I expect deliberate coding with AI assistance to beat vibe-coded speed once companies feel the complexity bills land.

So yes, still become a programmer, and build the writing and business and architecting muscles alongside the typing one, since fundamentals like writing good code and controlling complexity will stay valuable while the tooling keeps churning underneath.

If I missed anything about surviving AI as a junior please tweet, I am @troysk704 and post a comment. We will fix it!