🤣

I would, but Matrix is being Matrix and I’m unable to join.

OK, Sable is fucking amazing. there’s like three other people on the FAF matrix that are using it now, thanks for the recommendation!

I’ll check it out

I would probably go with one of the ones that supports clustering. if you have that many users, you may very well end up expanding and needing the room to do so. ejabbard has a focus on scalability, and they’ve been around since 2002.

yes they do

XMPP is a pretty wide umbrella, with a lot of very mature alternatives that are stable and consume very little resources.

fair enough. but most of what we use matrix for is chatting, meme channels, and gaming. and the voice features on fluxer are better than matrix, especially since not all matrix clients even support voice.

I don’t need something “just like Matrix”. E2EE in Matrix sucks and we had to remake several rooms on the FAF Matrix because a bunch of people couldn’t read anything. I need something that actually works most of the time and doesn’t have constant and myriad problems. the fact that most of fedi uses Matrix is even more regrettable considering the plethora of mature alternatives that work well and are stable.

ugh matrix sucks and I wish it wasn’t the preferred form of communication for everyone on fedi.

once fluxxer gets federation I’m gonna be done with matrix forever.

seconded. the Flow Z13 with 128gb of unified memory is a beast for both gaming and local AI.

Since you didn’t list a budget, allow me to recommend the ASUS Flow Z13. That’s what I have, and if you get the model with 128gb of unified memory, you can safely dedicate 96gb of that to the GPU for use with local models, allowing you to very easily run models that are too large for people with two 5090s to run. technically you could use more, but you want to keep a fair amount restricted to use as RAM so the system doesn’t run out of memory. best of all, since the machine overclocked has a maximum of 90w for power usage, even under full load it uses less power than most people’s gaming rigs do when idle. I know from experience that it is possible to run this device off of a small solar panel. when you are doing something low power, like browsing fedi, the power usage drops to around 5-10w on cachyos, meaning you get 6-8 hours of use per charge for low-power tasks. that drops to 1 hour when gaming, since it has a 70wh battery and the system normally uses 70w under load, but like most gaming laptops you will likely be plugged in when gaming so you get the full performance power.

for gaming, it’s pretty good too. the AMD RYZEN AI MAX+ 395 in it has a Radeon™ 8060S built in, giving it about the performance of a 3060, except that it has potentially 96gb of VRAM. meaning you will have no issues setting textures to the absolute maximum in any game. I get good frames in most games, easily reaching over 100fps in most of the titles I play, like Dune: Awakening. you can use G-Helper for linux to do things like control fan and power profiles, if you want to really optimize things.

the biggest issue with this device is the price. it was never cheap, since it launched after the RAM shortage started, and things have obviously gotten worse since then. it’s a serious investment. but as far as laptops for use with local AI go, I don’t think there’s currently a better option, all things considered.

for browsing, use Waterfox with uBlock Origin. for apps and the rest of the system, use a VPN that blocks ads like Proton, or use a Pihole.

that’s not true, you can block them with a VPN. or if you have one set up, a Pihole works even better. I haven’t seen an ad in any android app in years.

thanks, California, but we didn’t give a shit what you thought in the first place

check out HaloFPX. I can dig up the URL if you can’t find it. it’s a runtime that’s made for Strix Halo that ships with custom tuned models for ROCmFPX (I think that’s what it’s called), and it can give you really good speeds.

one of the models it comes with is a custom Ornith 1.5 35B, and with HaloFPX I get speeds in Deepseek Harness that are so fast I can’t read fast enough to keep up with generation.

EDIT link: https://github.com/julianmb/halofpx

that being said, most apps won’t recognize the unified memory because because the Strix Halo is so new, so if you want accurate reporting of GPU usage, you have to set a specific amount. otherwise it will probably report 100% usage of 512M VRAM in apps like btop.

oh I know, but since I’m running on a Z13, there’s a custom runtime built for Strix Halo that makes the model a bit faster. and that runtime uses a fine-tuned version of Ornith that takes advantages of the specific architecture to increase t/s. I know, it’s already fast, but it’s nice to see those first tokens start generating within a couple seconds, and my agent crank out a few thousand tokens in less than ten seconds.

ornith isn’t good for coding, sure, but for automating the boring shit I don’t want to pay attention to, the model works beyond expectations. I do wish it was abliterated because one of the things I use it for is OSINT workflows and tracking down the personal info of fash, and the guardrails version doesn’t always like to do that, so the skills have to carefully word the prompts. I imagine it would get pissy if I asked it hard questions about China too, but I don’t chat with LLMs so I’m not sure.

there’s always a silver lining to every cloud!

I guess the sleuths were right. I’ve used local GLM models in the past, they aren’t bad at all. although I’m currently enjoying the speed and quality of Ornith 1.5 35b running on HaloFPX.

just wish there was an abliterated version.