i thought if it is written in python one can unpack the rpm and see the source code and only for apps using go or c/c++ source is needed. if there is something that is not interpreted and there are no sources available, it wasn’t on purpose.
Yes, you can also install the app and see the python code directly on the phone.
I guess that’s the sources for Whatsapp?
yes, thank you very much.
Exactly. That danger is obviously somewhat mitigated when you see a serious attempt to make the code a collaborative project and host it in the open. I go as far as to make my person known which means I have an incentive to take care.
That’s not the point. I don’t even want to download an rpm until I’ve looked a the code. If NO effort has been made to document it or even post links to a repo, I have no trust. I go through a LOT of effort to release apps so I take it as an insult.
EDIT: just to take the negative tone off:
I have a number of professional (independent) developer friends who use LLMs aggressively. They pay multiple providers hundreds a month. In the two cases in mind, what is most evident is the discipline. They are both documentation centered as a matter of course (clients) but also structure fanatics (one is a devops guru). I have seen what they can do (double the output). But, it’s a high bar. I’m suggesting that to do it properly with LLMs will be costly. But that doesn’t speak against using them. I’d ask:
- Proper public repositories with
- documentation
- tests (if automatic generation, automated tests, please)
Not that I live up to that ![]()
Personally, I’ve worked on Vision Encoder/Decoder models, for document processing, and further back in time on RNNs for musical composition. Things have come a long way, but, when coding, I’ve found the LLM approach slows me down. But I also haven’t tried since version chatgpt 4. I’ve had very good experience lately with PEOPLE using code I’m working on and helping make improvements. Or even radical rethinking. It also requires some management skills, but it’s a lot more satisfying and the solutions are much more interesting than what LLMs produce.
I’m not a real programmer but I work in IT and used AI in my work.
IMHO, using AI is as good or bad as using npm, pip to some other more or less verified lib in your code.
Every line you generate and incorporate into your project is a liability, even the ones you wrote yourself!
It saved me a lot of work in the setup and with unit tests but in general it’s as if you work with an apprentice and thus can be slower, then if done by your self in the beginning.
I would never accept AI generates slop as in a vibe coded black box that the “developer” doesn’t understand.
If you send in a project where you did worked with AI then it’s the same to me as if you used some libraries or used some dev help form fiver. Are there any policies for that already?
Using a dependency is definitely not the same as a vibe coded app.
In the first instance, the developers are trying their best to provide a good library. With vibe coding the authors mostly have no clue what they’re shipping.
Like the recent AI spam app that has a disclaimer that credentials are stored plaintext. The disclaimer is better than nothing, but this still a no-go and shouldn’t be published.
When I was creating BitSailor, I went to great lengths to patch even the official Bitwarden CLI because it allowed reading plaintext credentials, so I’m really not happy when someone vibecodes a half assed app that doesn’t even try to be secure.
Does this harbour-whatsapp actually work?
I’m guessing it’s just a different frontend to the web app? Like those electron desktop apps?
Does it reveal contacts and my own phone number to WA? (Just to clarify, that would be a big no-no for me)
Yes it does actually work, but its pretty rough. Really only messages between individuals work properly - in groups you get the messages but there is no identification as to who their from. The formatting is lousy with overlapping text in places and there are no notifications.
There appears to be no functionality for synchronizing contacts with your SFOS contacts db that I’ve found.
However, it is a native SFOS Whatsapp app, and maybe a good candidate for further ‘human’ development - so I wouldn’t automatically write it off just because some of it was created by ‘AI’.
Sounds interesting, but I’m terrified of touching it. The last time the SFOS community tried this was Mitakuuluu and that got a lot of people’s accounts terminated iirc
I think the EU has forced WhatsApp to make its service interoperable with other chat platforms, which would mean that in the not-too-distant future there may no longer be a need for a dedicated WhatsApp app.
Yeah, I’m banned to this day from any Meta platforms. ![]()
Lucky you!
………………..
This sounded quite unbelievable, and looking into it a bit more, it’s even more unbelievable: two proprietary “third party” applications made a deal with Facebook to be able to interact with WhatsApp. If, like me, you’ve never heard of either BirdyChat or Haiket, don’t worry, it’s because they don’t actually exist. BirdyChat has a massive 5.000 (no, there’s no missing zeroes) downloads on the Play store and had its first ever release not even a month prior to that announcement, in the middle of October of 2025. Haiket is even worse: it quite literally does not exist, and its website was launched three days prior to the announcement, on 11 November 2025.
Both of them are clearly strongly intertwined with Facebook, if not directly owned and controlled by them, created purely to pretend to adhere to the Digital Markets Act.
Edit: it’s even sillier: ‘BirdyChat’ was registered as a trademark in September 2025 by ‘Fyello Productivity SIA’. ‘Fyello’ was registered as a trademark in May 2025 by ‘SIA Birdy McBirdface’. It’s venture capitalists all the way down.
That will not fly with the EU if they don’t want to end up with a million Euro fine like Muskrat.
They’re probably just testing things with some small chat services before it’s rolled out to others.
They will, off course, try to drag their feet so it may still take a while before there’s real interoperability.
It depends on how you handle them.
Case in Point: npm left-pad incident - Wikipedia