Eh. Firefox is fine.
The only FF fork I’ve ever used for some time is Cachy Browser, as it shipped with my distro and was ostensibly amore optimized. But even they depreciated it in lieu of vanilla Firefox.
And Firefox gets faster security patches anyway.
I’m more interested in Chrome forks because it’s Google spyware. And, as much as I don’t like it, I find Chromium-based browser to be faster. That doesn’t matter so much on desktop, but the difference is pretty dramatic on Android.
Ungoogled Chromium does not support full uBlock Origin. Last I checked, it wont auto-update itself on Windows without a 3rd party tool, and I remember it having some other “quirks” from the stuff it strips out. The delay for security updates seems pretty minimal, too.
And personally, I like the bangs feature, now that I’m using Orion on iOS anyway.
But its based on ungoogled-chromium, so if you prefer to use upstream, that makes a lot of sense. Helium’s main pitch seems to be an “easier to install” ungoogled chromium anyway.
Orion syncs cross platform, DuckDuckGo does as well. And I believe you can sync with extensions.
Orion is mobile. So is Cromite.
DDG is pretty good too. I like its approach, with a UI that encourages whitelisting sites.
To those asking “which browser other than Firefox”
It’s fantastic. It’s Chrome, stripped of junk, with full (not lite) Ublock Origin natively supported and shipped. What more could you want?
And it can coexist alongside Firefox.
Cromite is also great, but its antifingerprinting is so hardcore it breaks some sites. That’s perfect for shopping/private browsing, but a bit much for daily driving unless tracking resistance is your #1 priority.
On iOS and OSX, Orion (from Kagi) is sublime. It’s Safari based (which you want for Apple stuff), but heavily modified with a native blocker, and supports extensions if you really need them. There aren’t many Safari “forks” like it.
I say this because I’ve been through a gauntlet of trying a bunch. Bromite, ungoogled chromium, waterfox, pale moon, Thorium, Vivaldi, all sorts of iOS apps and Firefox/Chromium forks. And these feel like endgame to me. Helium is just about perfect (as long as its development isn’t dropped), and Orion is close aside from some UI quirks.
Orion is great.
So is DuckDuckGo.
I don’t want to leap into your throat, but have you tried a clean install of a different distro on a USB? And I mean clean; no reusing your home partition, no weird configs until you test out-of-the-box settings.
One thing I’ve come to realize is that I have tons of cruft, workarounds, and configurations in my system that, to be blunt, screw up Nvidia + Wayland. And my install isn’t even that old.
Hunting them all down would take so long that I mind as well clean install CachyOS.
I haven’t bitten the bullet yet (as I just run Linux off my AMD IGP, which frees up CUDA VRAM anyway), but it’s feeling more urgent by the day.
I used to write a few things for HardOCP, and we always made fun of Tom’s for clickbait, chum, and straight up groveling to Nvidia. We straight up weren’t allowed to cite them, and had no desire to.
That was like a decade ago.
…Maybe they were alright when Tom was still with them, but that was a loooong time ago. Too long for me to even know.
Yeah. They’re like aging potatoes in a barn they need to use for something, and its an “answer” to the gaming GPU shortage without reallocating any silicon supply.
Kinda already done:
https://arxiv.org/abs/2504.07866
Huawei’s model splits the experts into 8 groups, routed so that each group always has the same number of experts active. This means that (on an 8 NPU server) intercommunication is minimized and the load is balanced.
There’s another big MoE (ERNIE? Don’t quote me) that ships with native 2-bit QAT, too. It’s basically explicitly made to cram into 8 gaming GPUs.
If you can get good results on gaming cards, then suddenly ordinary gaming hardware, run in parallel, may be quite capable of running the important models
I mean. I can run GLM 4.6 350B at 7 tokens/sec on a single 3090 + Ryzen CPU. With modest token convergence compared to the full model. Most can run GLM air and replace base tier ChatGPT.
Some businesses are already serving models split across cheap GPUs. It can be done, but its not turnkey like it is for NVLink HBM cards.
Honestly the only thing keeping OpenAI in place is name recognition, a timing lead, SEO/convenience and… hype. Basically inertia + anticompetitiveness. The tech to displace them is there, it’s just inaccessible and unknown.
I had an epiphany with this GFN push.
Nvidia H100s/B200s, what they mostly make for AI, can’t game.
But, they also repurpose some 4090/5090 silicon as budget “inference” AI cards, like the L40 and such.
My observation is that these “inference” cards aged like milk. No business wants them at scale, they’re just crap at running big MoEs, to the extent that Nvidia was doling out contracts like “Okay, if you buy this many H100s you have to take some L40s too”.
They have piles of these “low end” AI cards no one wants, so what do they do? Use them for game streaming.
Not trying to defend Nvidia’s predatory practices, but this does make a lot of sense. They’re essentially overstocked with cloud-only gaming GPUs already.
I’d do CachyOS. It’s a very new GPU. They even have packages specifically optimized for that CPU, and fixing stuff (other than simply rolling back) isn’t such a pain.
But I’m biased, as I like CachyOS.
Fedora if he’s not gaming.
Bazzite if he’s gaming. Or CachyOS.
I’ll give you the secret to easy linux: stick with defaults! Stick with distros aimed at whatever you’re tying to do, and you get a whole army of very experienced developers preconfiguring it all for you, for free. Instead of having to maintain breakage youself.
For example, do you want to learn all about debugging AMD drivers? Do you want to get into the intricacies of performant Proton setups, and environment variables, and kernels stuff?
You could just not, and get all that prepackaged!
Here’s just a sampling of some pre-configured stuff in my distro:
cachyos/protonplus 0.5.14-1
A simple Wine and Proton-based compatiblity tools manager for GNOME
cachyos/protontricks 1.13.1-1
Run Winetricks commands for Steam Play/Proton games among other common Wine features
cachyos/protonup-qt 2.14.0-1
Install and manage Proton-GE and Luxtorpeda for Steam and Wine-GE for Lutris
cachyos/umu-launcher 1.3.0-2
This is the Unified Launcher for Windows Games on Linux, to run Proton with fixes outside of Steam
cachyos/vkd3d-proton-mingw-git 3.0.r0.g6d97b022-1
Fork of VKD3D. Development branches for Protons Direct3D 12 implementation
cachyos-znver4/mesa-git 26.0.0_devel.216300.02cfc61cc93-1
an open-source implementation of the OpenGL specification, git version
Do I know a thing about how Proton works? Nope. Do I know anything about maintaining an upstream AMD driver for some kind of bug fix? Absolutely not. And I don’t have to! It’s just there, in sync with the rest of my system through some maintainer’s magic.
Try CachyOS or Bazzite and do not stray from defaults. If it suggests you use the legacy Nvidia driver, you say “yes sir” and do as it says.
A huge fraction of linux trouble comes from picking a distro with bad defaults, or messing with good defaults.
playnite
It seems they are planning to move: https://github.com/JosefNemec/Playnite/issues/59#issuecomment-3542246599
I’m planning to move to Linux in 2026 after P11 is done, and since Playnite is my personal blocker for the move, I will try to make some Linux version in 2026. Definitely not fully featured version on parity with Windows version of P11, and probably desktop mode only, but something daily drivable with Linux specific features (Wine/Proton integration for example).
If that goes well, I’ll start looking into P11 Avalonia porting proper say I mentioned in previous update.
The real issue is devs not wanting to pay for hosting server side anticheat. I
Or allowing self hosted servers. With actual mods that just ban people who are being jerks, and basic anticheat tools shipped to them.
Whatever the issue and solution, the current state of the gaming market still makes mass linux gaming kind of impossible. Not from the anticheat games specifically, but from the OEM problem.
But why?
Wouldn’t it be better to spend the same effort writing ffmpeg modules and interfaces in Rust?
keeping external dependencies to a minimum
This is… concerning, too.
Media processing code is difficult. It’s not even a pure coding problem, and often involves human perception, extensive, expensive experimentation and esoteric, buggy hardware APIs . Hence the whole point of ffmpeg is basically integration of external libraries, with immense amounts of labor already put into each.
There are some Rust libraries they could pull in though. I guess it’d be reasonable to focus on newer formats/codecs that have Rust implementations already, and let ffmpeg handle weird legacy formats.
Gaming is the best commercial inroad I’ve seen, and exploding.
But I think the “kernel anticheat” thing is going to be a hard wall until Valve works it out. Unfortunately, big OEMs don’t want to ship a “gaming PC” that can’t run Fortnite.
Mmmmm. Pure “prompt engineering” feels soulless to me. And you have zero control over the endpoint, so changes on their end can break your prompt at any time.
Messing with logprobs and raw completion syntax was fun, but the US proprietary models took that away. Even sampling is kind of restricted now, and primitive compared to what’s been developed in open source.
If you think of LLMs as an extra teammate, there’s no fun in managing them either. Nurturing the personal growth of an LLM is an obvious waste of time. Micromanaging them, watching to preempt slop and derailment, is frustrating and rage-inducing.
Finetuning LLMs for niche tasks is fun. It’s explorative, creative, cumulitive, and scratches a ‘must optimize’ part of my brain. It feels like you’re actually building and personalizing something, and teaches you how they work and where they fail, like making any good program or tool. It feels you’re part of a niche ‘old internet’ hacking community, not in the maw of Big Tech.
Using proprietary LLMs over APIs is indeed soul crushing. IMO this is why devs who have to use LLMs should strive to run finetunable, open weights models where they work, even if they aren’t as good as Claude Code.
But I think most don’t know they exist. Or had a terrible experience with terrible ollama defaults, hence assume that must be what the open model ecosystem is like.
No Linux client for that either, though it seems to be planned.