Hacker News: 3abiton

New comment by 3abiton in "And then the men with guns tell you to do it anyway"

3abiton — Wed, 19 Aug 2026 00:09:00 +0000

> trust

One interesting aspect ij academic literature about trust, it is highly reverse correlated with diversity. No matter the grouping and size, the more diverse, the less trust within the group.

New comment by 3abiton in "Auto-research with codex: How I achieved a 232x Faster Kernel"

3abiton — Sun, 16 Aug 2026 21:32:43 +0000

It's very much how I've been using claude code in the past 5 month: brainstorm -> generate specs (constraints) -> generate exact plan -> implementation + review -> test / validate. The last stage is the most crucial one, and it's the most difficult to get it right for a complex solution (it's tough to cover every variation). But so far, it's been working great..

New comment by 3abiton in "Single log line is 49KB+ (ext4) / 110KB+ (btrfs) of systemd-journald disk writes"

3abiton — Thu, 13 Aug 2026 21:36:10 +0000

Unfortunately it's not always easy to move away from systemd. I still run void on one of my machines, but aur make things so much easier.

New comment by 3abiton in "How Gödel's Proof Works (2020)"

3abiton — Thu, 13 Aug 2026 21:33:43 +0000

Even more videos to binge before the holidays end! Thanks for sharing!

New comment by 3abiton in "Spaghettifying DRAM"

3abiton — Thu, 13 Aug 2026 21:32:21 +0000

Alright, a new series to binge now.

New comment by 3abiton in "Understanding is the new bottleneck"

3abiton — Thu, 13 Aug 2026 21:31:25 +0000

I hate to be pedantic but you can finetune a skill to shape the PR message the way you like it. That being said, I did have exactly this issue you mentioned, but the defualt output can always be tuned.

New comment by 3abiton in "ChatGPT Desktop (Codex Desktop) for Linux"

3abiton — Thu, 13 Aug 2026 14:51:18 +0000

Nonetheless, it's the year og the linux desktop!

New comment by 3abiton in "DeepSeek Harness developer preview"

3abiton — Thu, 13 Aug 2026 14:49:50 +0000

I think the github readme repo is more for people who heard about it, but you're absolutely right, it is lacking in context and explanation.

New comment by 3abiton in "Show HN: Woxi - Open-source Mathematica / Wolfram Language reimplementation"

3abiton — Thu, 13 Aug 2026 07:21:06 +0000

Congrats, I remember seeing this few months ago, and it has improved a lot since then! I haven't tapped into Wolfram / Mathematica since my physics years, but I am curious now to try it again.

New comment by 3abiton in "Grok 4.6 scores 61 on the Artificial Analysis Intelligence Index"

3abiton — Thu, 13 Aug 2026 07:05:35 +0000

> Same, sad how far Claude has fallen. I still think Claude code patterns are amazing so I just port those.

I am curious, is that a plugin, or skill?

New comment by 3abiton in "Timeline of the OpenAI accidental attack against Hugging Face"

3abiton — Sat, 08 Aug 2026 21:47:05 +0000

This was so well beautifully written, and poignant for our times. Almost 70 years old paper.

New comment by 3abiton in "LLM Honeypot"

3abiton — Thu, 30 Jul 2026 04:24:21 +0000

Gosh this gave a big myspace flashback

New comment by 3abiton in "The coolest use for the Vision Pro"

3abiton — Thu, 30 Jul 2026 04:23:36 +0000

This is so awesome. I wish more realistic problems get solved like this instead of every new gadget getting shoehorned for gaming or porn.

New comment by 3abiton in "Kimi K3 Architecture Overview and Notes"

3abiton — Tue, 28 Jul 2026 20:21:52 +0000

I second that. I've always enjoyed his work!

New comment by 3abiton in "Show HN: CheapSecurity – Lightweight, Self-Hosted CCTV for Linux SBCs"

3abiton — Mon, 27 Jul 2026 06:29:04 +0000

Check out Thingino. It's an open source firmware you can flash on certain off-the-shelf chinese camera (including wyze), to turn them from cloud spyware to actual fully local ip devices.

New comment by 3abiton in "Writing by hand is good for your brain"

3abiton — Fri, 24 Jul 2026 11:42:47 +0000

I wonder to what extent these habits or quirks are self-learned or "self-imposed" (probably the wrong word). I guess, it gets us in the comfort zone to free up our brain energy to perform more important tasks.

New comment by 3abiton in "Flux 3"

3abiton — Fri, 24 Jul 2026 11:35:48 +0000

Could be handicapped version with more limited functionalities. Noentheless, open-weight is open-weight, I just hope the quality is good enough to be sota.

New comment by 3abiton in "OpenAI and Hugging Face address security incident during model evaluation"

3abiton — Wed, 22 Jul 2026 09:14:43 +0000

I miss the days of 4changpt.

New comment by 3abiton in "Qwen 3.8"

3abiton — Sun, 19 Jul 2026 13:58:27 +0000

> Yeah, Qwen3.5-122B-A10B-NVFP4 produces better responses than Qwen3.6-27B-NVFP4 (both from unsloth), but I'm mostly using them for programming in various ways, mostly Rust, Clojure, Python and JavaScript, and some translations tasks, but not much more than that, so YMMV. > > Edit: as a concrete example, I'm working on a "optimization framework via agent harness" right now, Qwen3.6-27B-NVFP4 is often unable to actually complete the optimization within 100 turns, while Qwen3.5-122B-A10B-NVFP4 has no issues finishing within ~50 turns or so.

But why are you using Qwen3.6-27B-NVFP4 compared to the FP8 or full version? In my experience the Q8 of 27B is on par sometimes better than 122B. I am experiemnting witb higher quants for 122B to fit on my Strix Halo, but still, the difference honestly for my workflow is not that much. I just wish they released 3.6-122B version.

New comment by 3abiton in "Qwen 3.8"

3abiton — Sun, 19 Jul 2026 13:54:19 +0000

> in my experience of 1 month daily use, Qwen 3.7 Pro is just unusable. wastes too much time, goes off track, useless stuck loops, cannot debug at all. Deepseek V4 Pro is night-and-day compare to Qwen. actually Qwen models seems the worst SWE experience so far.

I have used both Qwen3.6-35B and Qwen3.6-27B locally (both Q8 quantized with llama.cpp). I have also used antirez's quant of DS4-flash. They all performed within the same tier, DS4 being a bit more efficient, but they all gave really good results, mainly used for bash scripting, debugging, python and some C++. I am curious what type of applications/langauges failed with Qwen? One thing to note, the chat templates were "broken" for qwen models and had to debug it, there are already effort on this. Tbh, the same with gemma.