Not sure what discussion @PolatBirdal is having with himself. But @lutwolf definitely gets what i’m after.
This is going in the wrong direction… That wasn’t my goal in this discussion.
I know, I just got annoyed by him putting on this act of being formal. It’s not really any different from trying to put others down, so I just reminded him of that.
The problem is mostly that (appsCount+n) != appsQuality(+n) - more “fast code” doesn’t improve the quality but clutters the store rosters. One superior app beats 20 mediocre ones - Google Play Store is a perfect example of messy masses that do not contribute in any way to better usability.
Vibecoding enables everyone to “create” apps, but there a big flaws that prevent those from becoming must-haves.
AI can make specific repetitive tasks less dragging for the dev, but it should not yet create full programs.
Cheers.
An aspect I would like to consider is whether our tools are actually able to handle the results of vibecoding.
Years ago I could use a bit of time and actually look through all of harbour and storeman, gaining an overview of the available applications and possibilities, occasionally downloading out of curiosity to try some things, even though admittedly understanding very little of what I was seeing.
In the recent weeks, I found some creators (among which JimKnopfIoT took a prominent role) were basically flooding the repos with new apps, at a rate that I believe nobody can actually oversee. As @attah has stated, nobody really knows these apps, probably nobody will be able or willing to maintain them. (it’s also beyond my understanding how @poetaster manages all his apps, but he undoubtedly does, i guess a well-designed app doesn’t need a huge lot if maintenance)
I daresay the available mechanisms on our platforms to sort, organise and find are already borderline insufficient. I see, and fully understand, users stumbling around not finding the adequate solutions, while some apps are already around in dupli- and multiplicity next to others that are long outdated and obsolete.
How is that system of app distribution supposed to continue to function?
Strict separation? Tagging? AI agent caretaker?
*edit: uff- the thread grew quickly while I was writing…
Yes, if we are going to get to that point now, let’s ask this question: why are you missing this point? You talked as if we already have stores overflowing with native apps like the Apple Store, and on top of that AI coders are coming in and starting to make applications, and then we are saying these words to them.
I want you to know that there are maybe 6, okay let’s say 10 developers who are truly competent at what they do. Even these people are already only focusing on their own applications and taking care of their own responsibilities. As for the rest of the applications, most of them have been left abandoned directly after the years around 2023.
So, in short, our ecosystem is extremely poor.
Look, here I agree with you and Attah: low-quality applications are a thing, right? Yes, I’m not denying that. When AI writes code, it doesn’t put its pants on and pull up the zipper properly, so to speak. But I am someone with average Linux knowledge, I’m not a person who is a master at coding.
Let me explain why I reacted a bit harshly toward Attah. When Attah talks about these things, he is very good and very right in terms of being formal and professional, but I want you to know this: the screen recording application I made does not work badly in any way. Yes, I made this with AI, but I made it by carefully examining and refining it.
They put a “hack” label on me. Do you believe me? I even started doubting myself, thinking, “Did I do something I don’t know about? These people are calling it a hack.” I went and shared the open source code and had knowledgeable people check it.
Let me tell you this: nothing even remotely like what they claimed came out of it.
So now, what are these people who are acting all formal and professional here actually talking nonsense about? This is exactly why it drives me crazy.
Of course, that’s true
That’s the point I wanted to make: I’m not sure if the person will be capable of that, if he/she doesn’t understand a line of code.
Of course that’s a completely different question and not what I wanted to say.
IMO saying JimKnopfIoT is a problem because, what seems to me, he publishes the apps he wants to use so he builds them themselves (with or without AI), dedicating his own time to help making SFOS usable for people like me who’s using his Matrix client every day, because other ones were not fitting to my use cases, is just unfair.
Have I bought Jolla Phone 3 years ago when LLMs where quite shit in programming tasks, I’d sell it after second day, it would be totally unusable for what I need from a phone.
Yes, new developers here use LLMs, why? Because learning new framework / language is a major PITA. With LLM you can quickly explore possibilities, write some code, and if you’re seasoned developer in other ecosystem it’s not that hard to review it all properly and test. My harness often does some stupid stuff, it’s my job to tell it what is wrong, how to refactor it and make it better.
Documentation for SFOS is also not that good, my LLM harness is able to fetch sources of some libraries and find public methods and properties that have zero results on SFOS docs.
It took me 10 minutes today to find what cacheBuffer does in QT5.x’s List because this is not there when you search on QT docs (they only give results for 6.x), I had to search in Ecosia with explicit 5.6 to find some relevant docs.
Yet the LLM could tell me what it does and how to use it (I was merely trying to verify it during my review process).
The time when people write code is largely behind us. I respect people who refuse using LLMs just like I respect people who use VIM instead of Jetbrains or Zed. Everyone should pick their own tools and if stuff is open source, anyone can peak inside and check if this is written well enough. Be it here or at my daily work. The developer picks their tools, as a lead, I only check the results.
Sorry everyone, but this common bashing practice here by the “seasoned SFOS devs” reminds me the “real programmers” in 2010 who were bashing PHP as shit language for non-programmers, or mocking Wordpress developers back then as idiots who can’t write their own CMS.
Time is changing, nobody forces you to use Claude or Opencode. Chill out and let others do their thing.
I still have to defend JimKnopfIOT regarding the spamming on Openrepos. He’s had the apps on GitHub for quite some time. However, he didn’t publish them on Openrepos until later.
Yes, I worked on design for around 1.5 years. That’s why, for example, I ordered a Qt book. My goal is not to memorize everything, but to understand the general concepts.
And I also want to add this: when I had Linux knowledge on my PC and started doing things here, I gradually began learning the meanings of the topics being discussed, as well as many of the services in coding and the general way things work.
So I know the kind of people you are talking about. They tell someone, “Make me an application,” and completely leave it to them. However, the person should actually examine everything step by step. That is what I was trying to talk about as well.
My point is about separating the people with “diamond ideas” from the coal, while still respecting the idea behind the topic title. That is why I want to encourage developers. I even want to encourage someone who doesn’t know anything yet but has enthusiasm and persistence.
edit: But the general people here and the people from the previous discussion are not encouraging people at all; on the contrary, they are literally leaning toward pushing them away. And while doing that, they are trying to do it by intimidating them in a way. That is exactly the point where I lose my temper and get fed up.
I have to admit that when I read the thread title I couldn’t actually imagine it going any other way . . .
My intention was not to identify JimKnopfIoT as a problem in any way. He just caught my eyes as so many apps appeared within a short time.
I’m trying to advocate thinking about better systems to organize and make available and findable what is there.
And although I’m cautious, I don’t want to place a verdict of any kind, about AI-based coding, as I’m well aware that I know hardly anything.
I am sure the app development/maintenance process for SFOS feels a lot like the Wild West, very much because people have the freedom to code what they want. It is not structured like some “official Jolla process”, with JIRA tickets, Confluence documentation and official GitLab Pipelines with their respective development branches (let alone adequate testing runs, both automated and by test engineers). It is more like coding on your own, deploying sources and modules into OR and “testing in production” aka “the users test it”. It would benice if there were at least some basic workflow rules (proper labeling, tagging and documentation), but that’s something the community will have to let grow together.
Cheers
Ask this question yourself: Do you want to use software for secure communication that was generated by someone who is not a developer? How can you trust such a program?
Please stop making me loosing time.
I spent 10 minutes searching what I should answer to that.
Didn’t find.
Very unpleasant to read.
Indeed I skipped over that before…
@PolatBirdal: may I urgently advise you remove that deeply inadequate remark. Under no circumstance are blunt insults acceptable.
The browser situation fits in well here.
When I started using SFOS 4.6 I wasn’t able to sign in to any native matrix app. Because sso.fachschaften.org (needed for my matrix ID) wouldn’t load on the then-already ancient Firefox 78.
After the internal WebView was updated to Firefox 91, it briefly worked. When I switched to a Xperia 10 III in August '25 however, because Sailtrix stores the login as a Secret, I had to sign in again.
Firefox 91 was too old again!
So I thought, I have to wait until SFOS 5.2 releases so I maybe get a time window to sign in again. But then, Xmatic came. A matrix client which can use any browser for signing in! On a different device or the now-available native Firefox 153. Finally it broke the cycle.