Oh! Thanks!

We could start a buy/sell/swap space using existing Lemmy technology, if we left the messy details for users to figure out for themselves.

Messy details add up quick, though. Payments, currencies, mailing addresses, reputation, conflict resolution, returns, scammers and spam.

But “Craigs List” solved very little of that on day one, and went on to be pretty successful.

So it’s possible. happening now, I’ve just learned. Neat!. https://lemmy.sdf.org/c/buyselltrade

(Edit: Added link to the surreal thing. Thanks!) (Edit 2: I swallowed the onion whole. Mostly jokes, there, I think. I’m gonna leave the link anyway.)

These resources helped me successfully switch from Ollama to Lama.cpp. Thanks!

I plan to operate on expired certificates for two to four years, forcing all three of my users to click past a terrible warning.

Then I will automate my six day certificate rotation with Ansible (officially), but maybe actually with a stupid bash script in CRON (unofficially).

The real value is in the agility needed to update certs every six days.

Six days is still too long to leave a compromised key in play, but anyone who can rotate their keys every six days can probably also rotate their keys quite quickly if there’s a comprise.

I will study these. Thank you!

I’m not the OP, but I’m trying to figure out if local models are always painfully slow or if I’m missing something obvious in my tuning.

Stable Diffusion can whip up a picture in less time on the same hardware, than lama.cpp takes to decide to call an MCP function.

It seems like I must be missing something in my lama.cpp setup, but none of the guides I’ve read have clued me in to what I’ve done wrong.

Ollama performs similarly poorly on the same harsware, so I’ve probably managed to make the se mistake(s) at least twice.

Anyway, that’s the main thing I’m reading along for. Trying to increase my understanding until I catch my own mistakes.

Doesn’t outlook support imap?

Officially, it can. Many organizations turn IMAP off “for security reasons.”

Edit: Microsoft routinely turns off features for security reasons, so this comment may age poorly, if not already.

My experience with ethically run (if unethically created) local models has been similar. Lots odd hardware to get slow mediocre answers.

But the biggest reason AI competes with plain old web search is that they’ve enshittified the web search available to us.

Corporate assholes are breaking search to force people to use their centralized harmful AI answer guessers.

We can fight back with Searx. https://wiki.debian.org/FreedomBox/Manual/Searx

I’m so tired of “Ai big number” headlines. 100+ packages - but how many good ones? (Genuinely, I don’t know, as I don’t the recognize those listed.)

We need projects that diligently keep the slop out. We don’t need every protect to do so, but it is critical that some do.

We have Meta and Microsoft to provide quick-to-market low quality shit.

The version of Linux that matters is the slow thoughtful good software.

I’m not mad at slop slingers, aa long as they don’t secretly inject slop risk where not welcome.

Edit: Thanks for responses.

For anyone who has read this far, I encourage seeing responses below for informative and nuanced thoughts on this.

I also noticed how much Bill Gates didn’t address his ties to Epstein here. Of course, addressing those ties probably wouldn’t help with my suspicion that he’s a rapist abuser.

Bad llm code is just as bad as bad human code.

Bad LLM code is worse than bad human code. LLMs sanitize out markers that humans use to catch and correct mistakes.

Eye contact is benign for most of us. We still give it a context warning in civilized circles. It costs us almost nothing, and lets people make an informed decision what you engage with.

I run Linux at a workplace that is mostly Windows and some Mac.

Nobody notices or cares.

Everything I need runs in a browser now.

But why does everyone bring up AutoCAD? Lemmy seems to be full of AutoCAD professionals.

I sling code and documents. Linux is better for code and uses Office 365 as well as any other OS.

• Agentic AI: A list of traditional IF THEN statements that do things based on the output of the magical LLM and then create a new prompt that they feed back into the LLM in a never ending loop, with the aim of consuming as many tokens as possible.

It’s so much worse than that. We use the pleasant answer predictor to predict what function calls and arguments passed to those functions might be the most pleasing.

It can provide all of the capabilities of piping curl into sudo with none of the warm fuzzy feeling of knowing I’ve turned over control of my computer’s fate to a stranger on the Internet.

Instead, I’ve turned my computer’s fate over to a statistical approximation of the slurry from mixing together the outputs of every stranger from the Internet.

Exactly. And they need to make back the biggest up-front loss in the history of investment, and probably - even adjusted for inflation - in the history of currency.

And all that on a product that people don’t want.

“This is not an AI problem. It’s negligence on a 40-year-old standard—and it’s basically every sci-fi film ever,” says longtime security and compliance consultant Davi Ottenheimer. “‘Highly isolated’ and ‘escaped through the one hole we left open’ cannot both be true.”

Oh, yeah. None of my usual techniques would help get newer KDE on stable.

I guess there’s a chance it’s in back-ports, or will get there soon.

I can’t comment on Debian Testing.

I just run stable and pull cutting edge stuff from back-ports or flathub.

For some fancy stuff there’s usually a .deb file to download from the official site.

If I’m really desperate, there’s always piping a curl command to sudo, if I really really trust the provider.

Vibrant! Captures both their beauty and their joy. Impressive!