Hacker News: loire280

New comment by loire280 in "We're releasing the weights of Mistral Large 4"

loire280 — Thu, 29 Oct 2026 19:44:15 +0000

It's better than nothing when you are troubleshooting and suspect it's their bug. I hope the rampant illegal AI scraping doesn't force companies to end the practice and go completely closed-source.

New comment by loire280 in "We're releasing the weights of Mistral Large 4"

loire280 — Thu, 29 Oct 2026 18:24:57 +0000

A well known example is Gitlab Enterprise - you can see the source code but (IANAL) you can't use it without paying them and any modifications you make become the property of Gitlab. I remember seeing similar terms around some of the proprietary Confluent Platform components.

https://gitlab.com/gitlab-org/gitlab/-/blob/master/ee/LICENS...

New comment by loire280 in "For the first time ever, Google Search is less than 50% of Alphabet's revenue"

loire280 — Thu, 29 Oct 2026 18:12:31 +0000

Google has been extremely aggressive in growing its Cloud business in the last few years and has been offering big incentives to companies willing to switch from other providers. Looks like that's beginning to pay off as companies start to pay the unsubsidized bills.

New comment by loire280 in "We're releasing the weights of Mistral Large 4"

loire280 — Thu, 29 Oct 2026 14:46:52 +0000

"Source available" is already a common term for source code that is provided with a very restrictive license or under regular copyright.

New comment by loire280 in "There isn't an easy exploit for this PHP vulnerability"

loire280 — Thu, 22 Oct 2026 03:38:40 +0000

The author isn't disputing that these organizations have developed working exploits, they're saying that the exploit is difficult enough to achieve that, as far as they can tell, nobody has publicly released a working exploit except for vibe coders whose exploits rely on insane misconfigurations on the server.

With previous showstopper security bugs, once you knew the vulnerability existed the exploit was immediately obvious.

New comment by loire280 in "Using an open model feels surprisingly good"

loire280 — Tue, 28 Jul 2026 03:19:20 +0000

Rolling your own Linux is nearly free and could be done on any computer you had lying around. Dropping >$5k on a computer to run a local model (badly) doesn't really scratch my "hacker" itch. The author of this article works for an AI infrastructure company and ran Kimi K3 on their infrastructure - this post reads like marketing.

New comment by loire280 in "Minecraft Java raises recommended memory to 16GB ahead of Vulkan transition"

loire280 — Mon, 27 Jul 2026 15:41:02 +0000

https://www.minecraft.net/en-us/article/minecraft-java-editi...

> For minimum requirements, the new target is 1080p at 30 FPS using the Fast preset. This means a 64-bit system, 8 GB of RAM with a discrete GPU or 12 GB with integrated graphics, a 4-core processor, and a Vulkan 1.3 capable GPU with at least 2 GB of VRAM.

> For recommended requirements, the new target is 1080p at 60 FPS using the Fancy preset. This includes 16 GB of RAM, a stronger modern processor, and a more capable graphics card with 6 GB of VRAM.

This blog post also implies that the requirements bump is due to numerous changes to Minecraft over the years, not just the change to Vulkan.

> If you already play Java Edition without issues, you may not notice anything changing right away. If your game often struggles, the new requirements may help explain why and give you a clearer target for a smoother experience.

New comment by loire280 in "CrankGPT"

loire280 — Mon, 15 Jun 2026 19:15:00 +0000

I agree this type of web design sucks. It's been common for more than a decade - I remember Apple getting criticized for using this on the product page for the old "trash can" Mac Pro in 2013, and it was already widely used back then.

However, it seems pretty clear to me they did this in service of a joke - you have to "crank" your scroll wheel to get to the content, just like you have to crank this device. I think it's funny...

New comment by loire280 in "Small models also found the vulnerabilities that Mythos found"

loire280 — Sat, 11 Apr 2026 17:27:47 +0000

> Anthropic's own scaffold is described in their technical post: launch a container, prompt the model to scan files, let it hypothesize and test, use ASan as a crash oracle, rank files by attack surface, run validation. That is very close to the kind of system we and others in the field have built, and we've demonstrated it with multiple model families, achieving our best results with models that are not Anthropic's. The value lies in the targeting, the iterative deepening, the validation, the triage, the maintainer trust. The public evidence so far does not suggest that these workflows must be coupled to one specific frontier model.

The argument in the article is that the framework to run and analyze the software being tested is doing most of the work in Anthropic's experiment, and that you can get similar results from other models when used in the same way.

New comment by loire280 in "81yo Dodgers fan can no longer get tickets because he doesn't have a smartphone"

loire280 — Mon, 06 Apr 2026 18:04:46 +0000

This happens to everyone's fingers to some extent because the fingertips dry out as you age. It's a huge source of frustration for elderly folks since it adds to the confusion around using touch interfaces. My family members have had some success moistening their fingers with a wet paper towel periodically as they use their devices, though of course that is impractical on the go.

New comment by loire280 in "Every layer of review makes you 10x slower"

loire280 — Tue, 17 Mar 2026 05:32:10 +0000

I've seen engineers I respect abandon this way of working as a team for the productivity promise of conjuring PRs with a coding agent. It blows away years of trust so quickly when you realize they stopped reviewing their own output.

New comment by loire280 in "AIs can't stop recommending nuclear strikes in war game simulations"

loire280 — Wed, 25 Feb 2026 21:17:52 +0000

Unless you're suggesting the toy company secretly rigs the magic 8 ball to never recommend nuclear war, I'll take my chances with the organizational changes.

New comment by loire280 in "GPT-5.2 derives a new result in theoretical physics"

loire280 — Fri, 13 Feb 2026 20:40:23 +0000

It's easy to fall into a negative mindset when there are legions of pointy haired bosses and bandwagoning CEOs who (wrongly) point at breakthroughs like this as justification for AI mandates or layoffs.

New comment by loire280 in "A shortage of tenors"

loire280 — Wed, 11 Feb 2026 19:35:50 +0000

I sing tenor in a university choir as an older male community member and we encourage anyone to sing tenor whose voice is low enough. Over 1/3 of our tenors are women and our voices blend very well IMO.

New comment by loire280 in "The Singularity will occur on a Tuesday"

loire280 — Wed, 11 Feb 2026 00:06:18 +0000

The Fremen followed a messianic figure into a galaxy-wide holy war because the Bene Gesserit seeded their culture with manufactured prophecy as a failsafe.

New comment by loire280 in "Yawning has an unexpected influence on the fluid inside your brain"

loire280 — Wed, 04 Feb 2026 21:19:55 +0000

Reminds me of a recent finding that attention lapses in a sleep-deprived brain correlate with flushing of cerebrospinal fluid (almost a garbage collection pause).

https://news.ycombinator.com/item?id=45771636

New comment by loire280 in "Voxtral Transcribe 2"

loire280 — Wed, 04 Feb 2026 21:07:34 +0000

They don't claim to support Polish, but they do support Russian.

> The model is natively multilingual, achieving strong transcription performance in 13 languages, including English, Chinese, Hindi, Spanish, Arabic, French, Portuguese, Russian, German, Japanese, Korean, Italian, and Dutch. With a 4B parameter footprint, it runs efficiently on edge devices, ensuring privacy and security for sensitive deployments.

I wonder how much having languages with the same roots (e.g. the romance languages in the list above or multiple Slavic languages) affects the parameter count and the training set. Do you need more training data to differentiate between multiple similar languages? How would swapping, for example, Hindi (fairly distinct from the other 12 supported languages) for Ukrainian and Polish (both share some roots with Russian) affect the parameter count?

New comment by loire280 in "No management needed: anti-patterns in early-stage engineering teams"

loire280 — Tue, 13 Jan 2026 21:41:28 +0000

I may not be using the same definition of "motivation" as the author, but understanding what motivates your people, putting the right mix of people together to work on the right problems, and knowing how and when to apply pressure to get people to do their best work are absolutely something managers can do to motivate their teams.

New comment by loire280 in "Eulogy for Dark Sky, a data visualization masterpiece (2023)"

loire280 — Sat, 10 Jan 2026 17:16:07 +0000

I still mourn the loss of Weatherspark's old Flash interface, which brilliantly displayed all of this data in a single pane to give context to the recent, current, and forecasted weather. I've never seen as concise a visualization of current and historical weather data.

New comment by loire280 in "How will the miracle happen today?"

loire280 — Fri, 09 Jan 2026 16:51:05 +0000

That's not the point of his piece, and spending time virtue signaling to the reader would undermine the message that this kindness is a form of grace, given freely without expectation of reciprocation.