To be fair, I didn’t say that at all. You should use all the tools at your disposal to get you where you want to go. If the tool exists and it can help you, it would be silly not to use it.
I was just trying to explain why some programmers (like myself) don’t want to deal with AI-generated code or donate my time to help those that use it.
(P.S. There are reasons not to use AI tools even if they are useful, such as the ethical aspects.)
It would be great, don’t get me wrong. A JUCE AI agent would certainly be beneficial to many people. I’m just saying you don’t need it to write code in JUCE or to convert Figma designs to JUCE.
A different way to think about this: it was mentioned earlier in the thread that it’s hard to get deep musical knowledge and it’s hard to get deep programming knowledge, and that’s right! I spent 7 years in music school, and then decades as a professional software developer.
When I read these threads, all I hear are people coming in and complaining that it sounds like too much work, and that I (or someone like me) should devote a lot of time building something that encapsulates what I’ve learned so they can use it to compete against me without any form of compensation to me. You’ll forgive me if I don’t think that’s a very generous offer. I’m happy to open source tools I’ve made for other developers, and wish I had more time for my own projects, but no thank you to this idea.
Also – anyone who insists that this is the only path forward and that we don’t or won’t need human programmers any more is too new to the field to have lived through any of the previous hype waves that made the same promise. Anyone remember CASE tools from the 90s? Round-tripping UML editors from the aughts that would let managers sketch stick-figure use case diagrams that would magically get turned into code?
The tools that are being developed will definitely impact the field and become part of the standard toolkit (I mean, I’m old enough to have once been lectured by a graybeard about how compilers were just a fad), but there’s an inherent and necessary complexity to the hard parts of software development.
I like how often I’m hearing “it’s here now so use it" instead of “this is useful”.
Using ai to get something that doesn’t work because you didn’t invest the decade in learning about the deep topics, and then being surprised that people who did invest that time don’t want to help you for free on a forum is a very entitled place to be. People don’t want to actively patch leaks in the thing that’s trying to make them obsolete without being paid for it. You can’t fault that. Plus these people tend to not be interested in learning, they just want the AI slop fixed and working. That’s why in a nutshell people are immediately turned off by these lines of questioning.
I’m not saying this is what you are doing, but you need to at least view the current landscape through the eyes of others and realise your position will likely be shutdown for the above reasons, and it’s entirely rational.
Exactly. The entire premise of “I’ll use AI to generate solutions to topics I don’t have deep knowledge about“ is deeply flawed because then how are you supposed to know when the AI’s output is incorrect? And now you’re trying to fix / maintain code that you don’t understand.
Really I view these AI code generators as more like technical debt generators. They may provide some short-term momentum, but I’d be willing to bet that in the long term, the overall cost to eventually fix that code is greater than it would’ve been to develop it correctly in the first place. I believe it’s easier, faster & cheaper to write a greenfield project from scratch using good software engineering practices from day 1 than it is to take millions of lines of legacy code with no docs, no tests, and lots of bugs and clean it up to meet those same standards.
Would you bet on it? 
If you have sufficient knowledge about prompt engineering and know how to use it, this stops being a problem.
From what you said here it doesn’t sound like it stops being a problem - if all you need is “sufficient knowledge about prompt engineering” then why do you need an LLM specifically tuned on JUCE? My 2c are that rather than relying on someone else making the dream solution for you, your time would be better served learning the fundamentals of how to use JUCE so you have the toolset to solve these problems yourself
While I agree it’s easy to make a big messy codebase with modern AI tools, they are actually extremely useful when treated as what they are: very fast and clever but sometimes stupid coding assistants.
Like any tool, it takes time to learn how to use them well.
You absolutely need to know your domain well in order to use these tools well. It’s a bit like hiring a hot-shot junior programmer and saying “build me X” - you need to work very closely with them, regularly check their output and make clear (and evolving) plans. There are times when they will save you a lot of time. And there will also be times when they cost you time (like making a wrong assumption that you didn’t notice, or they miss something you would have seen had you been working on the code yourself).
Basically, when using these tools your role becomes more of a planner and code-reviewer than a writer of code.
I’d rather write code though ¯\_(ツ)_/¯
I also saw that study…
I’m not surprised, LOTS of code checking and bug fixing is also not my idea of a creative process. It can also be quite frustrating.
I will say it’s quite good at algorithm classes though, like filters, where it can interface with Juce buffers, even the 20 gig downloaded A.I.s know about the Juce framework.
BUT it can’t ’create’ anything, certainly not in an iterative manner.
Don’t forget, AI is the worst it will ever be. 
People used to say that typing to take notes is less effective than writing.
Whether these studies suffer from sampling bias is a completely ignored, and it’s an easy topic for labs that want to put papers out.
The generalisation of tool usage ignores outliers and benefits the tool manufacturer 
I find this intriging. Maybe it’s time for AI poetry to rule our hearts.
PS. Not for me. AI for me is like an intern who screws up constantly because of not being able to see the big picture.
