Featured image of post The Wall of AI Development

The Wall of AI Development

The Wall of AI Development

〜Made it to the 8th station, but from there it was hell〜

Hello, I’m kenji.

Recently, I’ve been hearing a ton of people saying, “I asked AI and it made an app for me!” We’ve moved past the era of no-code and low-code, and are entering a phase of “vibe coding” (writing code by feel).

For example, if you say, “Make an app that processes images and posts them to SNS,” AI will spit out the code and even the UI. It’s amazing, you’d think we don’t need humans anymore, right?

But you know, that’s like landing at the 8th station of Mount Fuji in sandals.


Getting to the 8th station looks “easy”

Development with AI is really easy at first. File I/O? It runs for now. Network? Well, it connects. Database? JSON is fine. UI? ChatGPT spit out React code. Billing? Just copy-paste the Stripe API.

Here, you get the illusion of “Am I an engineer now?”. But the real hell begins from here.

img.png


Why you get stuck

The reason is simple: “AI makes it look like it’s done, but leaves the final adjustments entirely to humans.”

  • Data gets lost in File I/O
  • DB normalization is weak so searches are slow
  • UI is not intuitive
  • Crashes from access concentration because no load testing was done
  • Rejected by Apple/Google review
  • (Almost) causes an uproar by violating legal requirements

In other words, **AI is good at creating a prototype that looks like a finished product , but ** finishing it into a product that actually works in the real world is still a human’s job today.


But there’s a paradox

Here you notice one paradox.

What if AI becomes able to truly do everything in the future? In other words, what if “AI can perfectly handle security, billing, and design all by itself”?

At that time, the need for humanity to create apps itself might disappear.

Because if a user directly tells the AI, “I want to do this kind of thing,” the task will be executed immediately without an app.


🌀For example

In the past, we “launched a calculator app to calculate”, but now it’s done just by saying, “Hey Siri, what’s 12 x 32?”.

In the same way, if you say “AI, process this photo and share it”, the process might be completed without going through any app UI or API.

In other words, when the goal of “making apps with AI” is realized, a world where apps themselves are unnecessary might arrive.


In the end, what we can do now

So, what should we do?

  • Right now, “above the 8th station” is still the human domain, so polish that
  • Focus on essential user understanding and service design
  • Think about “what’s valuable” rather than “what can be made”
  • Rather, explore ways of providing value that are “not bound by the format of an app”

AI is a tool, a lift, and sometimes a competitor. But, “what to make” and “why to make it” remain questions for us humans.


Conclusion:

The wall of AI development is not a paradox of technology, but of structure

An era where anyone can get to the 8th station of Mount Fuji. But the real thing starts from there.

And by the time we reach the summit, a future might be waiting where we ask ourselves, “Did we even need to climb the mountain?”.

img_1.png

comments powered by Disqus