panda_abyss

Between NextCloud and Wordpress, this is a headache.

Not that NextCloud is a bad, it’s just widely deployed by people who are t strictly following releases.

I immediately thought of Hurley from lost when I saw them

They can’t check if it’s parched because they can’t use their browsers

The certificates were determined by fair die roll

I can’t believe the private key was 4.

I don’t have experience nor do I want to get into it, but openclaw is a very token heavy application and you can probably do a lot better performance wise.

The reason it takes so long for first message is the amount of prompt openclaw requires. It’s almost certainly overkill, and with low prompt processing it will spend most of its time reading memory and soul files and config info for all of its tools.

I think you could just install earandil pi, set a prompt, and add a memory plugin and this will be faster. This is what openclaw uses if you strip back all the extras.

Also look for mixture of experts, they’re far less GPU intensive, so you’ll get better mileage. No idea what models are out there, but like even an a3b or a4b will be slow but maybe usable.

Again, I do not want to get into what this use case lacks or doesn’t work for, that’s your business and I don’t want to get into it. I’m just saying openclaw is probably like driving a 12 ton Humvee to the grocery store.

Well, if it’s paramount I can’t watch it, so I’ll enjoy my middle aged Picard

If you update the model json you can pass through almost any parameters you want

It is an extra step though

I pointed lemonade at my llama.cpp model folder but it uses json files to configure them. I think it was fine for importing them or you could have an agent write them for you. After that you can manage models inside the app

Overall it wasn’t a huge deal, it’s easier than managing llama.cpp directly with rocm updates.

I run lemonade on my strix halo box, I change the models I use, but the qwens are pretty good.

I access it via Tailscale on my phone, then I have a page I can use. The Lemonade interface forces you into the app, so you’ll want some other interface for your LLMs, as the app is junk.

I used to run llama.cpp and llama swap, but lemonade combines them and makes managing different ai backends easier.

Archwiki is pretty good, it has a lot of breadth, but it’s not a manual and it’s not easy.

Sorry, that guide was written for systems running systemd and the install manual you followed was a from an opinionated Arch user for SysVinit.

Oh, this guide is 11 years old written by someone who hated Wayland, so no, modern KDE won’t work with it.

While I acknowledge that this chart exists, I do not subscribe to following it.

There’s no way to know when my annual compliance training will come up, so yes, I will automate it with puppeteer and Jev.

Just a heads up, I pulled the drive stats from Backblaze’s drive survey, for most brands failure rates accelerate at 4-5 years.

At that point Hitachi, Toshiba, and Seagate all increased in failures. Western Digital didn’t, except for their 2014 batch that immediately dropped off.

Western Digital post 2020 are the most reliable drives.

Hitachi, Toshiba and Seagate seem to hit a cliff around 2-5 years now, but that only kills about 10% of their drives, so with replication you’re probably fine, just make sure to mix drives so you don’t get correlated failures.

Bug fixes are good

No problem.

The other thing with AI stuff is it conveys very little trust. It’s easy-come easy-go, so it’s hard to tell how much maintenance the owner will has put in or will put into the project.

Scheduling is high risk, so it’s important that it feels safe

I hope this doesn’t come off as too harsh, but the text reads as very AI generated, and I think if you rewrote the main text you could do a much better job.

It’s not clear to me why to use it, the pitch today is that it makes failures loud, which is nice, but very poorly described.

The tool itself looks good, but the website looks very “Claude”, resistant the text at the top saying “ONE BINARY · LINUX · MACOS · WSL · 25 MB RAM IDLE”

Those are great pros, but are they the guest thing the user should see?

Honestly the first paragraph of the quick start is a much better pitch:

RunWisp is one small Go binary. You list the shell commands you want to run in a runwisp.toml, and RunWisp takes care of the rest — firing jobs on cron, restarting services when they crash, and capturing every line of stdout and stderr so you can go back and look at what happened. There’s a TUI in your terminal and a Web UI in your browser for poking around. No external database, no agent, no sidecar

This tool looks decent, but for me I find nix helps manage the complexity of systemd well

Gemini does the same style, I think mistral as well.

Oh, isn’t bitter that site curating a list of content for building AI models? I seem to recall that and hope it works except then from any laws, right?

I guess you’re watching Full Metal Alchemist Brotherhood

Full Metal Alchemist. Full metal alchemist.

Can I get “8-12 years ago” in your pool?

It’s been a while since I saw it, I remember the swimming scene where he’s looking at her but he acts perfectly fine

There’s a teenage girl who gets overly sexualized, but at the same time, it’s kind of so over the top it’s pointing out the issue that her other characteristics are ignored for it.

It’s also a film set in the 20s/30s in Italy and the Adriatic, made in the 80s, so kind of a product of the time. Apparently the dialog in other languages plays out slightly different, so the English dub might be part of the issue.

Personally, I do not think it’s over the line or gross, but some people do and it makes them not enjoy the movie.

Fantastic movie!

There is a problematic aspect, but it’s never too problematic to be a deal breaker.

You’re forgetting ancient Egypt and the Samurai.

I appreciate that Valve is trying and putting out good stuff

It’s just unfortunate everything is on fire while they try.

~100gb with images and the max image.

Asides from my kiwix clone of Wikipedia, in the apocalypse there’s not much value to most of the things I self host.

I self host backups, code forge, some AI tools (all my AI chat and completion are local now).

But realistically, in an apocalypse situation I’m going to leave my suburban home and migrate somewhere safer and more directly connected to food+shelter, and probably spend my days dealing with trying to survive.

My self hosting is primarily designed around avoiding US based tools and systems, so that I have more control over privacy and don’t find policy I disagree with.

In this case it’s the os asking your birthday, not having to prove your identity to use the computer

Which while still shitty is the only acceptable amount if she verification.

It’s still a fantastic computer.

I use it as a server and it’s very very fast, especially threaded workflows, and IO is fast.

Just don’t buy it expecting to replace paid AI services. And don’t buy it for AI dev, on paper it should be good, but driver issues. DGX Spark is better if yo want an AI dev machine.

It would not be worth it as a replacement for Claude.

80% of my issue is that it’s AMD and their drivers are still awful. 20% is that the token generation speed very slow, especially compared to commercial models running on dedicated hardware. MOE models are fine, dense models are too slow for meaningful workflows. ComfyUI is decent, but I’m not seriously into image gen.

I have a lot of fun with it, but I have not been able to use it for any actual AI dev.