AI Is Making People Feel Skilled Without Being Skilled
Access to answers is not the same as mastery. The risk isn't that AI does the work — it's that it does the thinking, and your skill atrophies before you notice.

I use AI every day. I'd be silly not to. Instant answers, code on demand, a tireless explainer for any topic I don't yet understand — that's a real shift, and pretending it isn't is a worse mistake than overhyping it.
But I've watched something happen, in myself and in the engineers around me, that we should be honest about.
AI is making people feel skilled without actually making them skilled.
Where the gap hides
The trick AI plays is subtle. You ask a question. You get a confident answer. The answer works. You ship.
You feel like the person who solved the problem. In reality, the solution moved through you without leaving anything behind. Next time the same problem shows up — or worse, a slight variation of it — you reach for the same prompt instead of the understanding you never built.
Multiply that across hundreds of small problems, and you end up fast at getting answers and slow at solving problems. Those are not the same skill.
What real skill looks like
Mastery has always been built in the part most people want to skip:
- The first time you don't understand something and have to sit with it
- The wrong attempt, then the second, then the third
- The afternoon spent reading a spec that finally makes a concept click
- The bug that teaches you how the system actually works, not how the docs describe it
That process is slow, frustrating, and looks like nothing is happening. It's also the only thing that creates depth. AI can shortcut the output. It cannot shortcut the rep.
The rule I use
I haven't stopped using AI. I've changed when I let it in. The rule is simple:
- Break the problem into smaller parts.
- Try to solve it myself first.
- Use AI only when I'm genuinely stuck — and then ask it to explain, not just do.
- Repeat the loop until I understand it well enough to solve the next variant without help.
It's slower in the moment. It's significantly faster over six months, because the depth compounds.
Two questions worth asking yourself
When you finish a piece of work, ask:
- Could I do this again tomorrow without the prompt?
- Could I explain why it works to a junior engineer?
If both answers are yes, you've gained skill. If either is no, you've gained output — and lost a rep you can't get back.
Use AI as a tool. Don't let it become the thing doing the thinking. The boring, slow, frustrating part is not a flaw in the process. It is the process.
Written by
Oben Desmond Ashu
Full-Stack Engineer · Social Finance UK
How I Price My Services as a Software Engineer
After 30+ projects across very different markets, here's the simple packaging structure I use to charge what I'm worth without scaring off the clients I still need.
This landed at the right moment. We're three months into a build we never validated. Bookmarking the prototype checklist.
The 'consistency over intensity' line is one I've been trying to articulate to my team for months. Stealing this.
Curious how you decide which prototype format to use — landing page vs Figma vs Notion. Is it gut feel or a checklist?