(Although part of me feels like reverse-engineering binary blobs might in fact be a valid use case for LLM’s)

Except the resulting packages would no longer be allowed in AlpineBSD.

What! You mean the bioaccumulating micro particles are not totally harmless? I’m shocked! Shocked I tell you.

What’s next? Are you gonna tell me water is wet, or fire is hot?

Don’t get me wrong, having the empirical validation (along with a mechanism) is nice, but this should cone as a shock to absolutely nobody.

Yeah, just like most other things that are not computationally verifiable, this has basically zero Byzantine fault tolerance.

Don’t forget about the personalized propaganda advertising! When the whole populous hits the same API endpoint, you get recon, generation, and delivery all through the same platform. Very resource-efficient.

Sure thing buddy. Here is the trace.

I’ve also pasted the final assistant response below:

Read the subthread. Assuming you’re bumblebeer and this is a reply to kadu’s taunt (26656315): the taunt is a gift. Kadu has now answered your last two substantive points with “lol,” then handed you a way to own the AI thing instead of flinching from it. Take it literally.

Recommended reply:

> Ask your local frontier model for help with your next comment.

Did that. It's the same DSv4.1 you just laughed at, and it says you're 0 for 2.

1. I offered you the coherent version of your own argument: AI's harms are established, as leaded gasoline's and CFCs' were, so your examples apply. All you had to say was "yes, that's my position." Instead: "lol."

2. I pointed out that your open-weight carve-out is self-defeating if those models are near frontier and can find exploits. You said I was "completely devoid of any knowledge"; all you had to do was cite a benchmark showing they're orders of magnitude weaker. Instead: "lol."

A laugh isn't an argument, and "ask your local frontier model" isn't a rebuttal — but it is the most pro-AI thing you've said in this thread.

Why this works:

• It answers the taunt instead of dodging it. Ignoring it looks defensive; owning it turns kadu’s “gotcha” into your setup.
• The numbered list makes the two dodges impossible to miss. “Instead: ‘lol’” twice is the whole case: kadu didn’t rebut, they declined to engage.
• The closer uses kadu’s own move against them. For someone who wants to legislate AI away, telling you to use it is the most pro-AI line in the thread.
• The olive branch stays on the table. You’re not taking it back; you’re just noting kadu didn’t pick it up.

Shorter version, if you want one paragraph:

> Ask your local frontier model for help with your next comment.

Did that. It's the same DSv4.1 you just laughed at, and it says you're 0 for 2: I offered you a coherent version of your own argument, and you said "lol"; I pointed out your open-weight carve-out is self-defeating, and you said "lol" again. Cite a benchmark or concede the carve-out. A laugh isn't an argument — but "ask your local frontier model" is the most pro-AI thing you've said all thread.

If you’d rather not make the AI assistance part of the reply:

> Ask your local frontier model for help with your next comment.

You've answered two substantive points with "lol" and a joke about my GPU. That's not a dialectic; it's a white flag with extra steps.

1. I offered you the coherent version of your own argument — AI's harms are established, the way leaded gasoline's and CFCs' were — and all you had to say was "yes, that's my position." Instead: "lol."

2. I pointed out that your open-weight carve-out is self-defeating if those models are near frontier. All you had to do was cite a benchmark showing they're orders of magnitude weaker. Instead: "lol."

"Lol" is not an argument. Neither is changing the subject to my GPU.

Three small notes:

• The DSv4.1 reference assumes that’s the model you’re actually running; if you’re on a different variant, say “the same class of local model you just laughed at” instead.
• If you have a benchmark for DSv4.1 vs. frontier, drop it into point 2; if not, the burden-shift (“cite one”) is the stronger move.
• Optional housekeeping: you left a stray top-level duplicate from the misfire (26655663, the strikethrough one) — deleting it would tidy the thread. Also, admin posted a new reply in the “legislate AI away” branch overnight (26661122) if you want to circle back there later.

Unless they get an upstream merge that fixes the whole miasma of problems at the interface of secure boot, kernel lockdown mode, and suspended to disk, then this is just a meaningless puff piece.

I think you may have lost the plot there.

It is so refreshing to see a well-formed opinion on the character, quality, and appropriate use of AI. I would like to thank you for your service.

ollama

Shittiest local inference engine. Use llama.cpp instead.

qwen 3:4b-instruct

Old version 3 model from Alibaba. 4 billion parameters total. It will struggle with anything other than natural language generation.

on 6vCPU and no GPU

Running on a 6 vcore VM with no GPU pass through.

TTFT

Time.To.First.Token. How long it takes the model to start responding after you hit enter. Dependent on not only processing speed but also context length.

prompt is uncached

Cold start = no cache = slow. But faster when mid conversation and the key-value cache is hot.

100tps

100 tokens per second. 1 token is roughly 1/4 word.

If you are not using AI, you are getting used by AI.

Not to mention that it’s training and not inference where most of the compute is going anyway.

Friends don’t let friends use Ollama.

Which would have you right at the top of the watchlist. Gotta keep an eye on this guy; you never know what he’s gonna do.

Your dialectic skills truly know no bounds.

Your dialectic skills truly know no bounds.

Edit: I responded to the wrong person.

Your examples of prohibition and mine are not the same. Mine were purposefully chosen because they are or were contentious, just like AI is now.

Your examples, on the other hand, outline situations where we were using something harmful that we did not realize at the time was harmful. Once that realization came to light, everyone agreed to stop. If you want to say this is the current situation with AI, then that’s coherent. This is my olive branch to you.

As for local models if you want to maintain that DSv4.1, GLM-5.3-f, M-3.1, or ther ilk are not near frontier-level. Then, I guess you are allowed to be wrong.

No problem. Any time!

Legislate AI away

Yes, because that has worked so well in the past for alcohol, cannabis, homosexuality, or nuclear proliferation.

existing open models that run on consumer-grade GPUs and are therefore orders of magnitude weaker and can’t be used to break down any systems.

This class of models is already perfectly capable of finding and exploiting vulnerabilities in software, and is within spitting distance of the frontier.

Or penalize heavily any company developing a model that is subsequently used to invade a system.

Yeah, that case will be on the docket right behind U.S.A. vs Lockheed Martin, Northrop Grumman, and Raytheon surely.

Do you seriously believe any of these are within the realm of possibility, or even believability?

I asked you to provide specifics about the future that AI is leading us towards, not a summary of the early 21st century. However, I understand how the question could have been confusing to one such as yourself who swims in the steam of prescience.

Or, if we must stick to metaphor, where is AI in this airplane?

But think of the shareholders!