I have been a strong believer in open source and free software for many many years, but if what I do is just used to feed Microsoft and boil the oceans I will seriously consider taking everything offline.
(That being said, in this particular case I didn’t actually create much of anything really. One might even say the process isn’t dissimilar to what an LLM would do, just slower.)
A little more on-topic: here’s some screenshots of GTK WebKit running on SailfishOS: (Note that the browser is NOT usable at the moment, but having the engine running is a first step):
So you are going the GTK way, or you plan to plug it back to a QML application ?
About Copilot and Gemini, the response it gives me tell me that every bit of code on the internet have been suck dry. Gemini was aware if the WebKit prototype for UB, that pretty underground info.
I second this, think it makes a big difference. My opinion is that it would be bad for sailfish if it was flooded with ai applications noone understands and can review
Awesome stuff, hopefully soon (two weeks?) AI will be good enough to add usbc-video-out too, who cares it’s slop, if it works it works and you can spend time reviewing the code to snuff out any deficiencies and reimplement it in a proper way after the fact, noone’s writing it themselves the old meatbag way anyway, just complaining jolla should waste their meatbag hours on this or that feature, so bring it on
If they work it would be great for sfos, all the community hurramphs (lets do the camera api2 work yay!) ended up with a sad violin as it’s too niche/specialist, if sloppa generator produced a working product who cares it was ai, you can spend months reviewing/auditing/tweaking the code later (just like people spent months ‘working’ on bringjng such features, oh wait they didn’t), or just don’t use it out of principles or whatever, the manpower is severely lacking, if people can deliver working features who cares what tool they used
Ah, not quite. The one big PR I’m dealing with took 24 hours of initial review. Now with a couple of weeks time between releaase, it’s evident that there is a lot of bug triage and dead code to weed out. In my case, this is ok, since it was a major bump, but I’m beginning to think I should not have accepted the PR but only cherry picked. Fire and forget with later cleanup could end up costing as much time as doing it from ‘meat space’.
The Jolla Browser use qtmozembed to integrate Embedlite that intergrate Gecko.
I replaced qtmozembed with WPEQt as the engine layer. WPEQt handles the WebKit process lifecycle and frame rendering, but I also wrote C++ bridge
classes to adapt sailfish-browser’s Gecko-centric QML/C++ API surface to WPEQt’s API.
My goal right now is to get a somewhat working Browser on my phone, then work backward to get proper RPM.
