Printed on: February 6, 2026
AI is altering how we construct apps. That is not information. What may be much less apparent is how this shift is forcing us to assume otherwise about what truly issues in improvement. On this submit, I need to share my balanced ideas on AI-driven coding. I might like to provide you my perspective on why tons of recent apps on the shop is not as scary because it might sound, and the way AI is making iteration cheaper than ever. I might additionally wish to discover how the human contact separates polished merchandise from vibe-coded slop when your pushed by offering a great UX and doing correct consumer analysis.
By the top, you will hopefully have the ability to make up your personal thoughts on how AI is altering what improvement means to us.
AI slop and the thought glut
Over the previous months, there’s been an enormous improve in apps being submitted to the App Retailer. Most of those apps are vibe-coded, constructed tremendous rapidly, and shipped to the shop as quickly as doable (and sometimes include a subscription).
To some, this proves that improvement is now simple. Anyone can convert their concepts into apps in a day. To me, it simply proves that concepts have been at all times low cost. Concepts have been by no means the differentiator of what makes an app nice. It was at all times execution.
Earlier than AI, anybody might have an thought for an app. The barrier was constructing it. Now AI has lowered that barrier considerably, so extra concepts are making it to the shop. However that does not change the elemental reality: a good suggestion poorly executed loses to a mediocre thought brilliantly executed. And let’s be trustworthy, in case you’ve been round in improvement lengthy sufficient you will know that the majority concepts aren’t actually that good anyway. Not less than not adequate to spend weeks of time constructing. With AI we will execute any thought in a matter of days or much less.
Should you ask me, what we’re seeing is a flood of apps constructed rapidly with out a lot thought put into the precise consumer expertise. AI will help you ship quick, however it may’t let you know whether or not what you are transport is definitely good. That also requires human judgment, style, and care. And extra importantly, it requires a target market.
AI makes iteration low cost
Here is the place issues get fascinating. Should you care about your product, have a target market, and need to make one thing that works extremely effectively, AI turns into a useful software for exploration.
You are now in a position to iterate sooner than ever. You’ll be able to take a look at extra concepts, discover extra choices, attempt extra prospects. And you are able to do all of this in your precise app earlier than you commit. The price of experimentation has dropped to close zero.
Let me offer you a concrete instance from Maxine, my health app. I lately added a streaks view. With out AI, this is able to have taken me a day or two to construct. With AI, I constructed it in about 10 minutes. That is basically a zero-cost funding to attempt one thing new and see the way it feels.
Now, I did need to make guide changes afterward. The AI obtained me 80% of the way in which there, however the final 20% wanted human consideration. And that is positive. That is the purpose. I ended up with a function I actually like, and it price me a fraction of the time it will have in any other case. To be completely trustworthy, this function would have been very low on my effort vs. worth checklist despite the fact that, as soon as I constructed it, it began feeling precious instantly.
I have been doing this consistently with Maxine’s UI. Between exercise classes (I work out 3 times per week), I am going to tweak the interface. At some point I am going to attempt a brand new format, get AI to alter issues, then truly use it throughout my subsequent exercise. If one thing feels off, I iterate once more. This fast cycle of build-try-refine is just doable as a result of AI has made the constructing half so low cost.
The code was by no means the laborious half. The code was, in some ways, the boring half. It was the factor we needed to do, the half the place we spent hours articulating our ideas and plans into precise implementation. Now AI handles that, and we will concentrate on what actually issues: constructing nice experiences and tweaking the small print to suit our targets.
Utilizing your personal app is the differentiator
Constructing quick is nice. However while you’re constructing quick and you are not utilizing the app you are constructing, you are not seeing what your customers see. You are not dwelling life together with your app. Usually that results in a reasonably common expertise.
AI typically will not actually do way more than common. It could actually generate affordable code (in case you’ve arrange good guard rails), it will make smart layouts, first rate flows. Nevertheless it does not know what it appears like to really use your app in the actual world. Solely you (and your customers) do.
Usually after I attempt a brand new UI in Maxine, I create one thing that appears fairly in isolation. It might need gotten some suggestions from friends primarily based on photographs, after which I truly use it throughout a exercise and I understand it is not fairly proper. Possibly it is too brilliant after I’m drained and sweaty. Possibly a button is simply too small after I’m dashing between units. Possibly the knowledge hierarchy is smart while you’re a screenshot however falls aside while you’re truly within the second.
Customers will let you know this too. They will say one thing is simply too brilliant, too huge, too small, no matter. And typically you possibly can’t fairly determine why till you set your self of their footwear.
This brings me to a UX precept I discovered a very long time in the past after I was in faculty. Think about you are constructing an app that helps customers examine practice occasions. Your consumer needs to know: can I nonetheless make that practice, or do I must catch the following one?
The intuition may be to construct an information-dense display. Present all of them their choices. 5 trains, departure occasions, platform numbers, switch data.
However take into consideration who’s utilizing this app. There is a good probability they’re in a rush. They could actually be operating towards the station, telephone in hand, attempting to determine the place to go. They do not have time to learn 5 choices and examine them.
So as a substitute of giving them all the pieces, give them the one choice that is in all probability what they want. Cut back steps. Make issues apparent. Your app won’t find yourself wanting as refined or complete while you’re sitting at your desk designing it. However when you begin utilizing it, as soon as you set your self in your consumer’s scenario, you will construct one thing that really serves them.
If you’d like your UI to be good for what the app does, you must use your app. It’s a must to construct area data. Typically you will construct one thing that does not look proper in isolation however works completely in context. That is not a bug. That is craft.
The human contact vs. vibe-coded slop
My solely actual argument in opposition to transport quick with AI is that this: do not ship quick for the sake of transport quick.
AI will help you construct slop actually rapidly. And you do not need to construct AI slop. You need to construct one thing good.
Positive, you may be tempted to chase fast releases for fast cash. However that does not make your apps higher. It does not construct one thing you are pleased with. It does not create one thing customers will truly love and follow.
The strongest argument in favor of AI-driven coding, and one I utterly agree with, is that it permits us to concentrate on what actually issues. The implementation particulars have been by no means the fascinating a part of constructing software program. The fascinating half is determining what to construct and the way it ought to really feel.
AI lets us iterate sooner than ever. It lets us attempt issues we would not have bothered attempting earlier than as a result of the fee was too excessive. However the human judgment, the style, the empathy to your customers—that is nonetheless solely on you.
If you mix quick iteration with real look after the product, you get one thing particular. If you mix quick iteration with indifference, you get slop.
Abstract
AI is altering improvement, however not in the way in which the doomsayers recommend. Sure, there are extra apps than ever. However concepts have been by no means the differentiator—execution was, and execution nonetheless requires human judgment.
The true alternative right here is that AI makes iteration almost free. You’ll be able to attempt issues, use them, refine them, and check out once more. The builders who will stand out are those who truly use their very own apps, who put themselves of their customers’ footwear, who care sufficient to shine past what AI generates.
Use AI to hurry up the boring elements. Use your personal judgment for all the pieces else. That human contact is what separates craft from slop.
