Hear me out: As we on Lemmy all know, there is the enshittification cycle. I’d argue that with LLMs/chatbots we are in the golden era. There is still a lot of VC money propping up unprofitable big tech ai firms, who then provide their LLMs for pretty cheap to get the consumers and advertisers hooked.
In a few years I assume that the prices will go up, maybe even the quality will go down if the ai companies dont find a solution for the ais eating their own shit problem.
So now is the best time to make AI products.
The whole copyright for code from small applications is already pretty over in my opinion.
So now: Lets build a lot of small tools/programms (I am thinking of GUI for ffmpeg, pixelart-editor and so on) with LLMs. In the future we can still improve them, either with or without LLMs.
(I am not really 100% of that opinion, but that thought came to mind and I am not really sure why it wouldnt be a good idea)
Writing code has never been an issue.
Non-devloppers imagine with LLM, creating program would become much simpler. Yet, it becomes simpler for non-developper. But simpler does not means simple.
There already was millions of skilled developers before LLM. But there is still no widespread ffmpeg UI, pixel art editor, …
There actually are dozen of tool like this. But as I said at the beginning, writing a program has never been an issue.
Creating a program that solve real problem. Build community of users (and contributors if it is open source). Gain trust of your users. Maintain it in the long run. Those are the real problem. And those are problem you face when you have written applications for years.
New “vibe coders” never faced those problem, so they are not even aware they exist. They imagine they can prompt their app and that’s it. But when you look at vibe coded apps in github, there were built one shot, by a single guy, and then just stay as is because no one use them. Not because they are bad, but because writing code has never been the issue.
Yup, I have a program (a programming language I wrote) I have been working actively for 2 years, and still add features now, all open source, with pages and pages of documentation, and I even added webapp. Still no users except me. I’ve published papers showing the program, and the analysis it can do. No users, no citations.
Now, someone with AI could probably make something similar, if they have a good amount of followers their program might get users too. And it might just be unmaintained after their interest wanes.
Hey, and you don’t even give a link to your project ? I want to see how it look like 🙂 I’ll probably never go back but still, I would like to give a look.
Use documentation is here, use the sidebar to navigate: https://nadi-system.github.io/latest/why-nadi.html
GitHub: https://github.com/Nadi-System/nadi-system
It has multiple components, all in the same github org if you wanna explore
Some example probects with networks and custom code. There is visualization too. You can run code line by line after loading network, or press Refresh all to run everything
https://nadi-system.github.io/playground.html?example
One of the example when loaded and run: (It’s better on computer)
Wow, impressive work. However, it’s a bit specific, and if you are not in the domain I can understand it may not seems interesting 😅
It makes me think a bit about Tulip, a tool developed by the LaBRI, French research lab about computer science: https://tulip.labri.fr/
I vibe code a lot. But I’m also 10+ years into professional level engineering.
My vibe coder friends are NOT experienced. My vibe coder colleagues produce piles of shit for me to fix.
I honestly think AI coding can be really good, but you need to read the source and understand the architecture. You also need you need to know how to engineer
I otherwise fully agree. AI is useful IF you are smart. And know is dirt cheap, so I build a bunch of projects I want knowing eventually I’ll have to code by hand once prices go up
No. If you can’t write code: Don’t.
No thanks
