Hacker News: cjbprime

New comment by cjbprime in "DeepSeek Harness Desktop for macOS and Windows"

cjbprime — Fri, 02 Oct 2026 05:04:42 +0000

It has particularly good observability (the ability to see the full content of every prompt and response and tool call) compared to other harnesses, that's the main thing that stood out to me.

New comment by cjbprime in "Astra for Coding: Why Are We Doing This Again?"

cjbprime — Fri, 11 Sep 2026 07:14:56 +0000

> 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like this one have been showing up constantly for the last three years, and literally all of them were obviously outdated and irrelevant within about a month.

New comment by cjbprime in "NTSB issues investigative update on B-767 runway excursion accident in Miami"

cjbprime — Fri, 11 Sep 2026 00:39:22 +0000

It could be literally automated. And I'm talking about while they were still minutes away on approach, not during final landing.

New comment by cjbprime in "NTSB issues investigative update on B-767 runway excursion accident in Miami"

cjbprime — Thu, 10 Sep 2026 23:25:36 +0000

And yet I am! Care to elaborate?

New comment by cjbprime in "NTSB issues investigative update on B-767 runway excursion accident in Miami"

cjbprime — Thu, 10 Sep 2026 22:52:55 +0000

I mean, complete ethical nihilism is one way to try to avoid losing this particular argument, but I don't think you'll get many people to agree with you on it.

New comment by cjbprime in "NTSB issues investigative update on B-767 runway excursion accident in Miami"

cjbprime — Thu, 10 Sep 2026 22:47:23 +0000

Maybe ATC could be expected to revoke landing clearance on extremely obviously unstable approaches? I wonder why that didn't happen.

New comment by cjbprime in "NTSB issues investigative update on B-767 runway excursion accident in Miami"

cjbprime — Thu, 10 Sep 2026 22:36:41 +0000

They made a series of at least ten (and that's kind of charitable) this-must-never-happen mistakes that likely rise to the level of extraordinary criminal negligence, including ignoring checklists, alarms, configuring the plane for landing in general, and basically everything related to safety.

You don't have to extend sympathy, just as you don't have to extend sympathy to drunk drivers who kill people.

New comment by cjbprime in "AI Responsibility – OpenAI and Anthropic"

cjbprime — Wed, 09 Sep 2026 04:06:05 +0000

> LLMs can't read binary directly without a disassembler.

They totally can. They're remarkably competent at disassembly.

New comment by cjbprime in "VMs won't contain cyber-capable agents"

cjbprime — Thu, 27 Aug 2026 00:47:15 +0000

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.

New comment by cjbprime in "The Commodore 77, an All-New Cyberpunk 2077 Collaboration"

cjbprime — Wed, 26 Aug 2026 03:21:30 +0000

If it was a mechanical keyboard with that case instead of an entire computer they could charge twice as much for it.

New comment by cjbprime in "VS Code in the Terminal"

cjbprime — Thu, 20 Aug 2026 07:22:24 +0000

Waaait what? That's both crazy and crazy useful.

New comment by cjbprime in "Memory prices climb 500% in 12 months"

cjbprime — Wed, 19 Aug 2026 10:03:21 +0000

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.

New comment by cjbprime in "Qwen 3.8 27B is excellent, but it defaults to overthinking things"

cjbprime — Mon, 17 Aug 2026 15:42:05 +0000

> 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?

New comment by cjbprime in "Qwen 3.8 27B"

cjbprime — Sat, 15 Aug 2026 07:04:30 +0000

> 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.

New comment by cjbprime in "GLM-5.3: Frontier coding with emergent cyber capabilities"

cjbprime — Sat, 15 Aug 2026 00:24:36 +0000

Why not? The web uses TLS, how's it different security-wise compared to a package download?

New comment by cjbprime in "Qwen 3.8 27B"

cjbprime — Fri, 14 Aug 2026 19:37:43 +0000

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.

New comment by cjbprime in "Qwen 3.8 27B"

cjbprime — Fri, 14 Aug 2026 18:11:58 +0000

Hm, I have a 4090 as well, and:

$ build/bin/llama-server -m Qwen3.8-27B-IQ4_NL.gguf --mmproj mmproj-BF16.gguf -c 170000 --parallel 1 -ngl -1 --cache-type-k q8_0 --cache-type-v q8_0 -b 1024 -ub 512 --flash-attn on --no-context-shift --no-mmproj-offload --spec-type draft-mtp --spec-draft-n-max 5 --spec-default --cache-type-k-draft q4_0 --cache-type-v-draft q4_0 --threads 24 --jinja --reasoning on -fit off

0.02.993.689 E ggml_backend_cuda_buffer_type_alloc_buffer: allocating 911.53 MiB on device 0: cudaMalloc failed: out of memory

Update: Oh, it works after I stop Xorg. But nvidia-smi only showed Xorg using 200M out of the 24G, so why would a 911M alloc fail?

New comment by cjbprime in "Run Kimi K3 using 29 GB of RAM at 0.50 tok/s"

cjbprime — Fri, 31 Jul 2026 17:29:23 +0000

Does it not use Metal, on macOS? Would it be faster if it did?

New comment by cjbprime in "Show HN: Getting GLM 5.2 running on my slow computer"

cjbprime — Fri, 10 Jul 2026 00:20:44 +0000

It's a very conservative warning. The application does not perform writes, so the application doesn't actually wear your SSD at all. The rest is just application-independent general hygiene.

New comment by cjbprime in "GLM-5.2 – How to Run Locally"

cjbprime — Wed, 24 Jun 2026 05:05:32 +0000

You still have a core misunderstanding. Only one layer of weights is required in memory at a time. A forward pass can be over-simplified as a matrix multiplication of each layer, one at a time.

There is no swapping of working RAM. We're just talking about loading the weights read-only data into RAM on-demand for each layer. It is only as slow as your storage interface.