Comment by linuxhansl | Hacker News Reader

Comment by linuxhansl | Hacker News Reader

Disclaimer: I use AI at work and for some projects at home and find it really useful (as long as you know when and how to use it)

With that said... It's funny how we went from "vibe coding" to review burden, sloppy contributions, copyright issues, unclear ownership (as in: who will fix bugs), needlessly verbose code, etc, etc. Several projects have banned AI contributions now and counting.

As I said, I use AI to help architect and write code and really like it; still, I cannot even count the times anymore where the model (we use multiple of the frontier ones) has completely gone down the primrose path. I always make sure I make it my code, every line is reviewed and understood, maybe I rewrite the AI output here and there, I understand the tests and testing strategy, etc, etc.

So I wonder where this is all going. We are past "vibe" coding already, and so far monetization of AI has been "patchy". On top of that local AI is becoming more and more viable.

The problem is primarily social. When code is cheap and entry barriers don't exist, how do you filter what is valuable?

For private projects, being able to code things quickly is a godsent. Also for professional environments with clear rules and mutual respect, this can be solved. But on the internet, with people you don't know? You must have some hard barriers.

I see 2 possible users for a coding AIs.

1. Professional coders. Us.

2. Non-professional users, people that need software to just get X done.

I am focusing on non-professionals.

If you look at phones you find apps to book flight, book hotels, exercise, meditate and whatever. Almost anything without a social aspect can be redone with AI.

Why redo with AI? Because it's personal.

I don't need 55 filters when renting an apartment on a business trips: i always want a terrace and good internet, show me price ranges, something close to where i will go to work. And after 10 times i use the app with AI, it should start to know what i like, and after 100 it should better do it.

I want my flights never to arrive in the evening because if it delays i don't want to be stuck in the airport.

I don't really exercise, but i really want to so my app does does need to have 80 exercises, it needs to have 2 basic ones and to use each dark pattern to make me do it.

I love movies, so i want to know when it's coming out.

Lots of this can be done with a simple chatbot, that goes out and find planes or whatever. Yes that is going to be ALSO how we use AI.

The distance between app and chatbot is somewhat of a grey area.

An app is a specific personal tool. An app has an interface, that makes it more reliable, it persists things i like, history and personalizations. An app "I built with AI" likely has some IKEA effect. And much more.

Overall, it's simpler, i know where buttons are because i told AI where to put it and they are big buttons, and if tomorrow i want my personal rent-a-car app to never ever propose company X that just gave me a shitty car, it will do it.

If i want to see the address in can ask AI, but i can also just open the app, no AI, no internet, etc.

I actually built some of those apps, for fun. We are not fully there for a non coder, but not too far.

In terms of security and code, etc, run them in a sandbox, do not put personal data, send to website for actual purchases/actions, zero inbound network connections ....

I am hand-waving, but i don't feel like i am bullshitting if say that the right security tradeoff will be found, no?

In any case I tried to give you an example of how coding AI can mean a sort of paradigm shift for software. I do not mean it in a prescriptive sense, "this is exactly what the future will look like", it's just feel ready to be surprised by how this will develop.

Reply on news.ycombinator.com