Hacker News: pulse7

New comment by pulse7 in "The Goose Programming Language"

pulse7 — Sat, 24 Oct 2026 20:03:40 +0000

Educate me: how to create a program in Goose which reads a JSON structure, prints it out and finishes. How is memory for JSON allocated on stack? It can be from 1K to 1M - depending on the file...

New comment by pulse7 in "Thelio Mira AI Linux Workstation: 192 GB GPU Memory"

pulse7 — Fri, 11 Sep 2026 08:16:32 +0000

Prompt processing on Macs is VERY slow...

New comment by pulse7 in "Training a 3.8B LLM to 0.384 CORE for $998"

pulse7 — Thu, 10 Sep 2026 08:34:46 +0000

I hope we will soon have an open-source project for training such small LLMs where one can just pick the architecture (like Qwen / DeepSeek / etc.), parameter count, dataset, ... and then let it run on a local/rented GPUs...

New comment by pulse7 in "IBM Unveils Next Generation Dual-Architecture Processor for IBM Z and LinuxONE"

pulse7 — Wed, 26 Aug 2026 21:36:25 +0000

"code translation in hardware" microcode is called "native"

New comment by pulse7 in "RISC-V: They Should Have Known Better"

pulse7 — Sun, 16 Aug 2026 08:20:25 +0000

So it's mostly the "Optionality". Like USB. And yet USB is everywhere...

New comment by pulse7 in "Qwen 3.8 27B"

pulse7 — Sat, 15 Aug 2026 05:00:02 +0000

It there anything similar for RTX 3090 and RTX 4090?

New comment by pulse7 in "Qwen 3.8 27B"

pulse7 — Sat, 15 Aug 2026 04:48:35 +0000

Can you please tell which Gemma 4 variant managed to correctly reason through your private benchmarks? Was is Gemma 4 31B?

What quantizations and context lengths did you use for Gemma 4 and Qwen 3.8 27B?

I am asking because I can't even load Gemma 4 31B on my GPU with any reasonable quantization (even with small context), while I can run Qwen 3.8 27B with large context and good quantization...

New comment by pulse7 in "Qwen 3.8 27B"

pulse7 — Fri, 14 Aug 2026 20:20:35 +0000

"--context-shift, --no-context-shift ... whether to use context shift on infinite text generation (default: disabled)"

From: https://github.com/ggml-org/llama.cpp/blob/master/tools/serv...

New comment by pulse7 in "GLM-5.3: Frontier coding with emergent cyber capabilities"

pulse7 — Fri, 14 Aug 2026 07:34:37 +0000

It will come... all big hardware players (Intel, AMD, Broadcom) and dozens of startups (Tenstorrent, etc.) are working on it...

New comment by pulse7 in "Docker Sandboxes – Disposable, isolated sandboxes for AI agents"

pulse7 — Mon, 10 Aug 2026 07:38:56 +0000

Hasn't Docker always been just a thin layer of duct tape over existing solutions?

New comment by pulse7 in "Why Wall Street is ignoring big tech's debt [video]"

pulse7 — Sun, 09 Aug 2026 22:08:44 +0000

You cannot wear out an SSD through AI inference alone. LLM weights are only read from the SSD, and read operations do not contribute to SSD wear. SSD wear is primarily caused by write and erase operations.

New comment by pulse7 in "Nvidia, Microsoft, Meta warn against overregulating open-weight models"

pulse7 — Fri, 24 Jul 2026 21:48:04 +0000

Those are routers - not general purpose computers.

New comment by pulse7 in "The day Steve Jobs dissed me in a keynote (2010)"

pulse7 — Fri, 24 Jul 2026 10:08:54 +0000

On the absolute limit of ethics? This is simply not ethical anymore...

New comment by pulse7 in "Quickly rewrite Git repository history"

pulse7 — Fri, 24 Jul 2026 09:57:59 +0000

The same here! Merged several repos into monorepo are retained full history!

New comment by pulse7 in "Bonsai 27B (1-bit LLM): The First 27B-Class Model to Run on a Phone"

pulse7 — Tue, 14 Jul 2026 19:31:25 +0000

Most probably not optimized yet for this model...

New comment by pulse7 in "Jim Keller's startup is building a factory to mass-produce small chip fabs"

pulse7 — Sun, 05 Jul 2026 21:06:18 +0000

With electron-beam lithography you can build transistors with gate lengths down to 1 to 3 nanometers.

New comment by pulse7 in "Jamesob's guide to running SOTA LLMs locally"

pulse7 — Fri, 03 Jul 2026 22:16:19 +0000

exactly: high demand!

New comment by pulse7 in "Jamesob's guide to running SOTA LLMs locally"

pulse7 — Fri, 03 Jul 2026 22:15:20 +0000

NVidia Spark is much slower (low memory bandwidth)!

New comment by pulse7 in "Jamesob's guide to running SOTA LLMs locally"

pulse7 — Fri, 03 Jul 2026 22:10:04 +0000

Can you please share you llama.cpp server parameters to turn on modern LLM sampling stack?

Docs [1] say that the top_n_sigma is already in the default sampler list: "(default: penalties;dry;top_n_sigma;top_k;typ_p;top_p;min_p;xtc;temperature)"

[1] https://github.com/ggml-org/llama.cpp/blob/master/tools/serv...

New comment by pulse7 in "Rocketlab acquires Iridium"

pulse7 — Mon, 29 Jun 2026 15:27:12 +0000

This https://en.wikipedia.org/wiki/Elon_Musk%27s_Tesla_Roadster ?