I didn't expect it to work like this.
I thought AI would just make me code faster — like autocomplete on steroids.
But after a few projects, it stopped feeling like "coding faster" and started feeling like I wasn't really coding at all. I am orchestrating.
I've shipped a bit of things here and there for my own usage. Nothing crazy, just some shit I used in my workflow.
It's usually:
- something I got annoyed doing manually
- something I wish existed (actually most of them existed elsewhere, but it's a PAID version so I'd rather build it on my own lol 😄)
- or something I just wanted to test for myself
Also, these projects force me into parts of the stack I wouldn't normally touch.
An "easy" app turns into:
- auth flows
- database decisions
- payments
- edge cases everywhere
You don't really get to stay shallow.
The Shift That Changed Everything
I think the biggest difference is this:
I stopped thinking in code. Instead, I think about the overall system design and can focus more on the business decisions I make instead of finding logic bugs, system bugs, etc.
Before, I would think:
"I need a function that does X"
Now it's more like:
"I want a system that behaves like this"
And then I orchestrate the AI to figure out most of the implementation.
Not perfectly — it still messes up a lot — but it gets surprisingly far.
At that point, my job becomes:
- checking if this even makes sense
- spotting where it might break
- tightening the edges — even AI helps me in discussing edge cases
Less typing, more judging.
What I've Learned (mostly the hard way)
A few things that actually made a difference for me:
1. Start from the surface
If I can't clearly describe what the user sees and does, the AI output is usually garbage.
Once that part is clear, the implementation gets way easier.
2. Don't trust the code (at all)
AI code looks right very often. And for some reason, they HARDCODE something they absolutely shouldn't LOL. Especially Manus! I lost faith in Manus since that time it was acquired by Meta lol 😄
Anything that touches:
- user input
- APIs
- databases
…will break in weird ways if you don't validate properly.
I learned this by shipping bugs. Repeatedly.
3. Architecture still matters (a lot)
AI is good at writing pieces.
It's not great at deciding:
- how things should be structured
- where state should live
- what should talk to what
If that part is messy, everything becomes painful very quickly. But that said, Opus is really good at it if you can instruct it properly and systematically.
4. Shipping fast beats overthinking
Some of my "bad" ideas only revealed themselves after I shipped them.
No amount of planning would've saved that.
Now I just:
- ship something basic
- watch what breaks
- fix from there
It's a lot less elegant, but way more effective. Anyway, I am the only user so far hahaha 😄
5. Every project should make the next one easier
This one surprised me.
Small things compound:
- scripts
- templates
- checks
- little internal tools
You reuse them without thinking, and suddenly you're much faster than before. Just like setting up Stripe and Redis databases — such a pain in the ass the first time, but it gets smoother, you know.
Across different stacks
One weird thing about AI — it doesn't really care what stack you're in.
I've bounced between:
- web apps
- iOS stuff
- backend scripts
- random automation
I'm definitely not "expert-level" in all of them.
But AI makes it possible to move across them… as long as you roughly understand what's going on. I think what's important is the basics and foundation of the things you work on.
The part people don't like to hear
This only works if you have some foundation.
If you don't understand:
- how systems fit together
- what "good" looks like
- where things usually break
…it's very hard to tell if the AI is helping you or quietly creating a mess.
I still read everything it generates. It's beautiful — honestly more enjoyable than reading stuff in my day job lol 😄
I still debug things I didn't write.
Sometimes I throw everything away and start again.
I'm still figuring this out
I don't think this is "the future of coding" or whatever.
I just know this is how I've been working recently — and it's the fastest I've ever gone from idea → something real.
It's less about typing faster now.
If there's one takeaway — stop thinking so much and just build. Stay in the flow, keep going, and let compounding carry you further than motivation ever will.
Happy building!