AI: "I'm Gonna Eat You"
2026-10-26
Thanks to AI, you can now build a pretty solid product even with no experience as a programmer.
Which raises a question. Say one person with a programming career and one without both use AI to build the same product.
Would there actually be a difference in how fast they finish, which tech stack they pick, or the quality of what they end up with?
Personally, I do not think there is even a guarantee that the product built by the experienced person comes out better.
To put the conclusion first: what is asked of programmers in the age of AI is not technical knowledge, but the breadth of view to notice what is needed right now.
So this time, from the hiring side, I want to write about the skills programmers will be expected to have from here on.
Experience barely shows until the product is done
Building to a spec is exactly what AI is best at.
In practice the spec is almost never clear from the start, but gaps like a broken screen flow or a missing message are things anyone notices once they build it and click around.
Now that AI keeps making development faster, I think in most cases it is more efficient to just get your hands moving and improve things in an agile way than to go back and forth pinning down the spec first.
All in all, I get the feeling that on the way to "done," whether you have experience or not hardly makes a difference anymore.
If anything, I think experience can even get in the way.
Take a programmer who clings to an old tech stack. They rewrite the code AI gives them into the style they are used to, or turn down the structure it suggests and build it the old way — and all that extra work turns straight into lost speed.
Someone without experience has nothing to compare against, so they can just go with what AI suggests.
I suspect this reversal, where experience holds you back, is quietly happening all over the place.
The hard part is running and maintaining it
On the other hand, I think there are still plenty of situations where experience really counts.
Someone without permission being able to read other people's data, costs shooting up as users grow, the app breaking on malicious input — none of these happen within the range a builder normally clicks through.
These gaps only show up after you release.
And once you are in the phase of running and maintaining the product, I think experience makes a big difference in how good it turns out.
To put it loosely, what matters is breadth of view.
- Wanting to tune the caching strategy
- Wanting to lock down Tailwind and shadcn conventions with a linter
- Wanting to expand the tests so the product meets certain requirements and performance targets
- Wanting to compare tech stacks to bring costs down
- Wanting to tighten validation, error handling, and RLS to protect against malicious third parties
My impression is that these are things today's AI rarely brings up on its own.
So if you are light on programming experience, I think there is a real chance you ship without ever noticing these gaps exist.
That said, fixing them is something AI can do with ease.
I already leave almost 100% of my coding to AI, but I still feel I have a lot to keep up with outside of implementation.
I suspect proposals and judgment calls like these are still a stretch for AI alone.
The important part here is that you do not necessarily need to know how things work under the hood.
Even if you know nothing about RLS, as long as you can notice "wait, could other people read this data?", AI will tell you where and how to protect it.
So what is asked of us humans is, needless to say, not the ability to implement, but the ability to propose what should be done before implementation even starts.
And whether you notice is exactly where I think experience starts to count.
The skills that are asked for
Seen that way, I think what you should ask in hiring changes too.
Rather than questions that test knowledge you could get from AI in no time, questions that test whether someone notices should show the difference much more clearly in this big age of AI.
For the same reason, I feel that take-home assignments are getting harder to use as a basis for judgment.
AI can write the code, and it can put together neat design docs and records as well.
So if I were on the hiring side, I think I would ask the person, right there, about judgment calls outside of implementation.
- Where in this service could other people's data end up visible?
- If you had 100 times the users you have now, what would break first?
- What would you put in CI, and why?
- How far would you go with tests?
- After releasing something you built, is there an improvement you added on your own?
More than whether the answer is right, I want to see where their attention goes and how far ahead they are imagining.
If you dig in on the spot, I think it naturally becomes clear whether it is a prepared answer or something they actually think.
Output is worth more than before
On the other hand, I think output like open source is now an even more useful thing for the hiring side to look at.
Building open source with AI is completely normal now, so that is not a problem.
With something like a coding assignment, everyone submits under the same conditions, so it is hard to tell people apart. Open source, though, is something nobody asked them to make.
What they decided to build, what they fixed after releasing it — that is where you can see what they have noticed along the way.
The same goes for blogs and other output beyond open source, and I think they make good starting points for digging deeper in an interview.
I feel it is getting harder and harder to stand out on implementation speed or the amount you know, thanks to AI.
What is left in the end might just be whether you can notice what is needed right now.
Even at the level of defining requirements is fine — I think what is wanted is someone who can figure out for themselves what is missing.
Flip that around, and I suspect people who are convinced they are right and never update their thinking will gradually get weeded out.
And I think AI is only going to make that trend stronger.