It is the question we now get in almost every enrolment conversation: if AI can write the code, why should I learn to?
It deserves a straight answer rather than a reassuring one — so here is what we actually think, having taught through the change and used the tools daily.
What AI genuinely does well
Let us not undersell it. Current tools are very good at:
- Boilerplate. Form markup, a standard layout, the fiftieth CRUD screen. Work that was always typing rather than thinking.
- Explaining unfamiliar code. Paste something confusing, get a decent explanation. For a learner, this is genuinely valuable.
- First drafts. A rough function that is 80% right and needs your judgment for the rest.
- Getting unstuck. The blank-page problem largely disappears.
Anyone claiming these tools are a gimmick has not used them seriously. A developer who uses them well is meaningfully faster than one who does not.
What it does not do
Here is the part that gets skipped in both the panicked and the promotional versions.
It does not know what to build
A client says they want "a website for my business". That is not a specification. Turning it into one — asking what they sell, who buys it, what they need a visitor to actually do — is most of the job, and it happens in conversation with a human who does not know what they want yet.
It does not know when it is wrong
This is the important one. AI produces confident, plausible, working-looking code that is subtly incorrect — a security hole, an edge case, a query that is fine with 100 rows and dies at 100,000.
Spotting that requires knowing what correct looks like. Which requires having learned to code.
AI raises the floor for people who already know what they are doing, and it is a trap for people who do not. The gap between those two groups is getting wider, not narrower.
It does not take responsibility
When a client's site goes down at 11pm, someone has to fix it. When payments fail silently for three days, someone has to find out why. That someone is paid for judgment and accountability — neither of which is something you can prompt for.
So what actually changed?
Our honest read: the floor rose and the ceiling rose with it.
Work that was purely typing is worth less than it was. If your plan was to make a living assembling simple template sites with no understanding underneath, that plan is in trouble — and it was already fragile before AI.
Meanwhile, what a competent developer can deliver in a week has gone up sharply. Clients feel that, and they expect more for the same money. The developers doing well right now are not the ones avoiding AI. They are the ones who understand their stack and use AI to move faster through it.
What this means if you are starting today
Learn the fundamentals anyway. Genuinely. Not out of nostalgia — because the fundamentals are what let you tell good AI output from bad, and that judgment is now the valuable part.
Then be deliberate about how you use the tools while learning:
- Write it yourself first, then compare. Ask AI after you have attempted it. The comparison teaches you something; skipping to the answer does not.
- Never ship code you cannot explain. If you cannot walk through what a line does, you do not yet own it — and you will not be able to fix it when it breaks.
- Use it to explain, not just to produce. "Why is this approach better than mine?" is a far more useful prompt than "write this for me".
That last one is really the whole discipline. Used as a tutor, AI accelerates learning. Used as a vending machine, it prevents it.
Why we teach AI as its own course
We built AI Starter — prompting, vibe coding and AI-assisted SEO — precisely because using these tools well is a skill, and using them badly is worse than not using them.
We are deliberate about the order, though. We do not recommend AI Starter as anyone's first course. Take it alongside or after Web Starter or Web Dev, when you can already read code well enough to catch it when it is wrong.
Common questions
Will AI replace web developers?
It is replacing specific tasks rather than the role. Boilerplate and first drafts are largely automated. Deciding what to build, judging whether generated code is correct, and taking responsibility when something breaks are not. Developers who use AI well are replacing developers who do not.
Is it still worth learning to code in 2026?
Yes, and arguably more than before. The fundamentals are what let you tell correct AI output from confidently wrong output. Without them you can generate code but not verify or fix it, which is the part employers and clients actually pay for.
Should I take AI Starter as my first course?
We would not recommend it. AI Starter makes a developer faster; it does not turn a non-developer into one. Take Web Starter or Web Dev first, or take AI Starter alongside one of them, so you can already read the code the tools produce.
Do you allow AI tools during the course?
Yes, with one rule: you must be able to explain any code you submit. Our courses are project-based and assessed on the work, and the point of the project is that you understand what you built.
The short version
AI has not removed the need to understand what you are building. It has raised what clients expect and widened the gap between developers who know their craft and those who were only ever assembling templates.
If you want to learn it properly — the fundamentals first, the tools alongside — have a look at the courses or come and talk to us.
