Hacker News: hrmon

New comment by hrmon in "DeepInfer: High-performance inference framework for DeepSeek models"

hrmon — Tue, 20 Oct 2026 15:58:49 +0000

US has only SI businesses.

New comment by hrmon in "Doory – Open-Source Canva"

hrmon — Mon, 12 Oct 2026 12:32:23 +0000

"yeah, that's what we did too."

New comment by hrmon in "GPT-6 Astra"

hrmon — Fri, 04 Sep 2026 20:33:18 +0000

On the introduction video: The lady asks "make sure that [the presentation] feels really high-end", but something that gets generated without any effort is not high-end anymore. it will be mediocre, or more probably GARBAGE.

New comment by hrmon in "Qwen3.8 27B scores 52 on Artificial Analysis"

hrmon — Mon, 17 Aug 2026 20:08:32 +0000

I want to highlight its (1-hallucation rate) at 70%. BRAVO! For me, this is its most wonderful score. GPT-5.6-Sol sits at 8%.

New comment by hrmon in "Models Are Getting Dumber on Purpose"

hrmon — Sun, 16 Aug 2026 20:24:04 +0000

A sharp critical thinker draws admiration, but it is a joy to talk to some dude who knows stuff.

New comment by hrmon in "Show HN: Maple-Preview – Ternary 20B MoE running at 120 tok/s on a iPhone"

hrmon — Wed, 05 Aug 2026 09:15:44 +0000

English Wikipedia: "As of 16 October 2024, the size of the current version including all articles compressed is about 24.7 GB without media."

Models in this size range should aim for correct tool calling and avoiding hallucinations, not universal knowledge. (apparently they don't.)

New comment by hrmon in "Seedance 2.5"

hrmon — Sun, 02 Aug 2026 09:04:04 +0000

In second video, Look at the reflection of the staff member headset :)

Such models will have fascinating use-cases. But "generating realistic content" is a comedic resemblance of realism movement in painting.

New comment by hrmon in "Web-based cryptography is always snake oil"

hrmon — Sun, 05 Jul 2026 10:07:49 +0000

Many comments suggest that the definition and criteria proposed are infeasible and useless. They are not wrong. But still it points out something the layperson misses (or even many tech-savvy people): An E2EE service from A does not provably protect your data against A.

New comment by hrmon in "The short leash AI coding method for beating Fable"

hrmon — Fri, 03 Jul 2026 06:31:41 +0000

But staff engineers take "responsibility"

New comment by hrmon in "GPT‑NL: a sovereign language model for the Netherlands"

hrmon — Wed, 17 Jun 2026 06:54:58 +0000

> This vaguely-nationalist world view around tech that’s emerging in Europe is dangerous, man.

Guess which country blocked access to a SOTA model based on national security bullshit.

New comment by hrmon in "A network smuggling Starlink tech into Iran to beat internet blackout"

hrmon — Sun, 03 May 2026 07:08:48 +0000

US military "tested" some of its new weapons during the last war on Iran, in one case killing more than 15 kids [1]. So US tech is famous for improving life quality in Iran.

[1] https://www.nytimes.com/2026/03/29/world/middleeast/us-preci...

New comment by hrmon in "Ask HN: What did you purchase that measurably improved your quality of life?"

hrmon — Mon, 26 Oct 2020 18:49:50 +0000

"Everything you own end up owning you!"

New comment by hrmon in "Ask HN: What is your favorite motto?"

hrmon — Fri, 18 Sep 2020 14:46:30 +0000

For God's sake that's the ultimate motto! :)) But really, I prefer one step in 20 directions because at the end I want to be a human being who has tried to be better in many dimensions of life, not the popular superhero at X who is a freak!