Is your argument that it would be really bad if it was true, therefore it's not going to be true? It would be really bad if I got cancer, but I don't respond to that by refusing to think rationally about it.
Possibly more along the lines of, if modern civilization collapses, it doesn't really matter if you buy, sell, or hold.
If anything, you should be stockpiling food and guns. Or, who knows, maybe GPUs and DRAM. Physical possession of goods, not paper ownership of the same.
Debt only exists by the mutual belief of debtors and debtees that the alternative would be worse.
> I’m more and more convinced that all of AI engineering is Neijuan (内卷, meaning curl inwards). In China it describes a system that demands ever more effort and competition without improving output
I don't know what to say, except that articles exactly like this one have been showing up constantly for the last three years, and literally all of them were obviously outdated and irrelevant within about a month.
The tokens spent to find the vulnerabilities will cost money, and the attackers will usually find themselves more financially incentivized to spend money on finding the vulnerabilities than the defenders.
When the "new" C64 was first announced I was really hoping they would add a physical switch on the back of it that would let you switch between using it as a computer and using it as a keyboard to drive your normal machine.
Some meta keys could get a bit wonky, but I'd snap one up in a heartbeat!
Yeah, I’ve actually used that keyboard before. It has a bit of a hollow feel. I really love 8BitDo’s game controllers, but I've always found their keyboards to be a bit lacking.
My other big issue is honestly that super "in-your-face" 8BitDo decal at the top. That’s the kind of thing that takes away from the whole C64 aesthetic for me, and it really should have been placed on the bottom.
I'd love to see Keychron do a C64 themed wireless 75% mechanical though.
I think they still sell the one from the previous owners that's an Atom-based PC in a beige breadbin. There are ways to use it to drive your workstation.
Not sure about that, V100s are all selling for less than $1k despite being 24-32GB VRAM around 1TB/s memory bandwidth, which is the same price point that 4090s and 5090s are commanding $4k-$5k for. Hardly hotcakes.
> strong recommendation: ignore that default. Run Qwen 3.8 27B on low or even no reasoning levels at first. It’s a great model, but wow that default setting is a bad place to start.
Has anyone tried asking the model to choose and emit the most appropriate reasoning level for each prompt, as the first part of answering it?
> General knowledge: I usually ask 2 questions many small models get wrong: summarize Operation Trojan Horse by John Keel, give publication year. Summarize the Ariel school incident of 1994. Qwen3.8 got the first question right, along correct publication year, gave glorious detail on Keel's theory, but got the second wrong. It thought that school was located in the USA. Ah well.
I'm not sure that this means anything. You're asking a ~27GB file to have losslessly compressed the entire training set (which apparently is a large chunk of the entire internet). That's not possible. Whether it happened to encode these particularly obscure facts losslessly or vaguely isn't really telling you anything about how good a model it is.
Like most models it makes up things when it doesn't know. I had high hopes for Gemma4, which was said to having 'solved' this particular problem - it didn't. Gemma4 made less things up, did say it didn't know more, but it's still far from perfect. By comparison Qwen3.8 knows more, but still makes things up when it doesn't know. Coding abilities are very impressive however, and yeah these SVG tests do seem to 'scale' or generalize over its general reasoning+coding abilities - at least in JS and Rust. My next test will be to ask it to write some macros in Racket, just to see if it can balance parens. Most, if not all models cannot, no matter their size.
Honestly it seems like a job for the harness, rather than the model. Sample the model with the same question, perhaps with varying temperature (?), and use that to establish a degree of confidence in the answer. If the model provides very different answers every time, respond that it doesn't know. If it responds with the same answer usually but a different answer sometimes, respond with moderate confidence. If the model always responds with the same answer, respond with certainty.
If we want to implement Intelligence, and especially now that "the box is open" we must, we can play with the "intuitive" LLM architecture to understand it and squeeze it to its potential yeld, but at some stage we have to actually implement intelligence. That implies notions of confidence and a Foundational Theory of Knowledge (knowing why you know something), among the rest (one shot learning, update through reflection etc.).
I... really don't agree. Being confident in using documentation to program in a variety of environmnets is not "terribly hacky engineering". Good engineering involves leveraging documentation well.
I actually disagree that this doesn’t mean anything. I understand the contention that it’s not measuring the quality of the model in general, but I think it is measuring something useful.
A good example of this is planning hardware projects - a larger 200b plus model like DeepSeek V4 flash will recommend parts like motors, real time clocks, voltage regulators etc and it will do so providing exact model names and specifications.
I wouldn’t expect a smaller model to encode all of this information, but it is helpful to understand where that cutoff is because it changes what the model might be useful for. It is a very crude way of measuring because it comes down to the balance of training data at sizes this small - but I do think it conveys something that is helpful in real world tasks.
Agreed. It's just to see how much knowledge the model still has for its size. For comparison, no other model of this size has ever given me the correct answer to the Keel question. Even GPT-OSS-120B got it wrong, but it's expected for its relative old age.
I concur, treating models as question and answer machines and judging them on recall is meaningless, unless you're measuring quantisation impact on a foundation model maybe.
100% agree, it's really just for kicks&giggles. The fact that the model answered correctly, unlike every other model of its size before it, still pleasantly surprised me.
It's less about malicious intent and more about predictability.
When I install software on my computer with apt, I trust that all the files will go to the right place and install scripts are going to do sane things relative to the rest of the system. And I can just uninstall the whole thing with one command later if I so choose.
If I curlpipe a script, I get none of those guarantees. I have seen curlpipes that put files in weird places, guess the wrong OS, and mess with config files that I didn't want them to touch. When they break or I want to uninstall, I have to sit down and understand a (possibly minified) script to clean things up manually.
Yes containers are a half solution to this, no I don't want to use containers 100% of the time.
Does anyone know how to get this working with Claude Code via llama-server? I'm getting a jinja template error about the system prompt not being the first message.
The person above is running the OS from a separate GPU, so the command can use all of the VRAM of the 4090.
If you're trying to use the 4090 for your desktop at the same time, you will need smaller context size. You could remove the '-c 170000' and try turning fit on to see what fits.
>But nvidia-smi only showed Xorg using 200M out of the 24G, so why would a 911M alloc fail?
That was just the last buffer allocation request that failed, it didn't tell you by how much it failed by. It could have failed it by a few kilobytes, it could have failed it by 910MB. One would guess it probably failed it by a couple hundred megabytes in the end judging by your results.
