What exciting things are likely to happen for Jolla and SailfishOS in 2024?
Are there any opportunities for us to pay money to Jolla to speed up new feature development and bug fixes in SailfishOS?
What exciting things are likely to happen for Jolla and SailfishOS in 2024?
Are there any opportunities for us to pay money to Jolla to speed up new feature development and bug fixes in SailfishOS?
Not necessarilly Jolla, but adjecent:
Edit: Altered Order
Qt6 and GCC update.
Please donāt make this just another wishlist-threadā¦ thatās not the question asked - and there is quite enough of those already.
Itās mistaken to hope for such things, but Iām hoping for peace in Europe and peace in the Near East.
After the moronic abandonment of Russian cooperation, the pace will probably return to what it was before.
The last blog entry is from February. That alone speaks volumes.
Itās snowing for the first time in seemingly ever so Iām looking forward to sledding. Maybe a bit of skiing.
Iām not sure what you mean? That over specified ārun this source in this webengineā thing? I mean, half of my bloody apps are just a web engine wrapper around some html. And yes, Iām preferring to keep the html src on the device and avoid the web.
Do you mean like āshortcutā to run a webpage in a container? The progressive part is a bit lost on me if you donāt port the dom stuff to QML.
Thatās the problem. The solution right now is to make an app thatās just a web container. ā Granted, having the technical know-how is nice, but having to make an app for every webapp isnāt as nice.
A shortcut could work, but I think having PWAs that allow for notifications, are theoretically standalone (having the browser UI invisible or similar), or are offline apps (true pwa), āinstallableā from the and/or a browser, is far more desireable than having a shortcut to open the browser (which disconnects every time you switch out of it) or having to make an app for it.
Closest thing we have is FratzenJail with custom URLs or lelig:web
āTogether with the Jolla community, we are going to develop and take Sailfish OS into the AI āāera. The new company has already agreed on more than EUR 10 million in funding for the implementation of the new strategy,ā continues Saarnio.
I have concerns about this. Whilst AI is obviously going to transform everything IT, itās difficult to predict how. The film Her required something close to AGI. The vast majority of the AI being plugged into everything at the moment is gimmicky.
I just read 31% of corporations are now using the likes of CodePilot. Itās possible now to train our own AI largely on Sailfish code eg OpenRepos though it does require a ChatGPT 4 subscription.
By the way, we are on the verge of ācodeless programmingā ie drawing a UI, explaining how itās supposed to work and the AI writing & packaging the whole thing for you. There is already a working demo of this creating simple programs using existing software.
All sounds very jolly (pun intended), but perhaps they should concentrate on fixing some of the longstanding basic functionality bugs in SFOS first?
On platforms like ubuntu touch you have numerous 'PWAās which are nothing more than a shortcut:
For instance,
[Desktop Entry]
Name=Ola
Comment= Ola cabs PWA
Type=Application
Icon=olacabs.svg
Exec=webapp-container --fullscreen --store-session-cookies --webappUrlPatterns=https?://book.olacabs.com/* https://book.olacabs.com
Terminal=false
X-Ubuntu-Touch=true
X-Ubuntu-Splash-Color=#7a8800
Which is nothing more than a standalone browser instance. I guess what youāre asking for is the webview version of sailfish-qml, so a runner?
That is obviously not an āinstallā at all. Still, this is a PWA on ubuntu touch. There are a bunch of others, many of them are, unsurprising, a QML webview wrapper around locally installed assets. Look, for instance at, Philipp Fruck / DartsTrainer Ā· GitLab (just an example). Thatās more or less how I build webview based things. But my approach is simple in comparison since it doesnāt require extra metadata (the PWA manifest stuff). But, that is a case of a ārealā install where all html/js/etc assets are installed locally.
Iām building a Model (VisionEncoderDecoder) for document analysis (ISO review processes). About 70% of the code is actually run of the mill algorithms. The āaiā, in this case, is mainly glue (classify, extract). I also run 35GB Codebooga (on cpu 128GB RAM). Now itās not as good as ChatGPT4+. But almost. And ChatGPT4+ doesnāt save me much time. Itās useful for one liners. OpenAI still manages to assert that 2 is and is not prime within a context window of 2048 tokens.
The real trick, in the end, is producing good UI. Thatās the most expensive part of building an app (or a game), in general.
Youāre a coder and very familiar with the Sailfish specifics. I basically havenāt touched coding since my CS degree 28 years ago!
And we have so few coders that this sort of thing could make a huge difference. AI can also document and explain someone elseās code.
It could surely update API calls too. You reckon Codebooga would be up to this?
I suspect Iād be good at UI design but Iām guessing youāre talking about the coding side.
It just says āthe AI eraā. A mere buzzword if you ask me. But itās nice to hear that cooperation with EU and ā¬10mil are on the table.
You mean like the WYSIWYG HTML editors of old? Those were a huge success!
Jolla has a lot of old code that needs to be updated. AI/KI software can provide great services there if there is a lack of staff. At our company, we also have a lot of technical debt and are looking for solutions. Amazon also has a lot on offer. Porting from Java 8 to 17 etcā¦ The question would be whether this would be possible with SaiflishOS. Effort and costs too, of course. OpenRewrite as an example.
There are a lot of misconceptions about what generative models can do. The first thing to note is that given a large enough context, ie. sufficiently detailed prompts to āguideā the AI, the number of iterations it takes to get functional output (never mind up to some āspecificationā) is high. And youāre back to a human programming. Where these machines excel is regurgitating RECALL (with the flair of a dishwasher) from a large corpus of previously programmed stuff. They canāt dream up something new. I can get codbebooga to make, for instance, a supercollider Synthdef (a synthesizer voice) that whistles. Well, it produces a Synthdef (that even works!). But it does not whistle. In any case, codebooga (and chatgpt) is good for ātemplateā stuff. And well known algos. But neither have been able deal with things like ISO specifications meets real world (multiple sheets of xcell) and you would need multiple tuned models. Iām working on one, that with a great deal of care may deal with 20-30% of the final workload. Maybe Iāll get good enough at programming MLs to get to 50%.
I trained as a Printer. I learned typesetting writing SGML on typesetting machines in the 1980s. Spent 10 years doing traditional and then Desktop Publishing, and graphic design. During that period I also, initially as a hoby, made programs. Games. So, Iām talking about the entire process. I have completely failed to build nice UI both at the hardware and software level in recent years. I guess Iām a bit pessimistic because Iāve just left the finer point of UX as an afterthought although I know how good it can be. I think, for instance, that @BlacksheepGER does really clear UIs in the Silica context.
Itās EU money??
You mean like the WYSIWYG HTML editors of old? Those were a huge success!
Dreamweaver is still decent as long as you donāt need efficient code (maybe they can get AI to update it). But this is a step beyond. No more fields in tables for eg CSS. And obviously, no real limits in terms of what can be created in the future. Thereās a video in the link from the /. page.
The only thing Iāve written with ChatGPT (3.5) is a browser extension. It took about 45 mins and I donāt know JS or Manifest files.