Hacker News: NorwegianDude

New comment by NorwegianDude in "Qwen 3.8 27B"

NorwegianDude — Fri, 14 Aug 2026 15:16:16 +0000

If the benchmarks are a real indication, we now have a local model that is runnable on a high-end personal PC that trades blows with the leading model Claude Opus 4.6 Max from half a year ago.

Insane if that is the case. Downloading now!

New comment by NorwegianDude in "llama.cpp"

NorwegianDude — Wed, 12 Aug 2026 10:24:51 +0000

llama.cpp has router mode these days, swapping out models for you, you no longer need llama-swap.

New comment by NorwegianDude in "Claude Code pricing: same tokens, same model, up to 40x the price"

NorwegianDude — Tue, 11 Aug 2026 10:51:35 +0000

But also remember that while the API cost might be >50x the subscription cost, the actual cost to run it is much much cheaper.

It will only become cheaper as models become more efficient.

New comment by NorwegianDude in "Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows"

NorwegianDude — Mon, 10 Aug 2026 15:18:04 +0000

That is the most recent Qwen and Google models, there is no newer version, yet. Qwen3.8 27B might come in a couple of days tho, if it's launched alongside the large one when the Qwen3.8 countdown reaches zero.

New comment by NorwegianDude in "Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows"

NorwegianDude — Mon, 10 Aug 2026 15:04:52 +0000

Well, I can run some models that are better than some of the weaker and cheaper Anthropic models locally, like Haiku 4.5, and solve tasks that would cost ~4500$ every day in tokens, so yeah, they are definitely extremely profitable on inference.

New comment by NorwegianDude in "DeepSeek V4 Flash 0731"

NorwegianDude — Fri, 07 Aug 2026 18:26:00 +0000

Eh, what are you guys even talking about? Deepseek is not cheapest provider as is, and it's MIT. So deepseek making it more expensive to use is just nonsense, they can only change their own pricing. It's the beauty of MIT license and open weights. If anything, these models are some of the safest in the world to use if you worry about a rug pull.

New comment by NorwegianDude in "Self-hosting Kimi K3: 20% more hardware cost, 20% better task resolution"

NorwegianDude — Wed, 29 Jul 2026 19:04:12 +0000

You can get easily over 100 tok/s on the Gemma 4 31B QAT if you enable MTP. Same goes for the 27B. I'm getting 880 tok/s of throughout on a single RTX 5090 for batch tasks.

New comment by NorwegianDude in "Kimi K3 exploited the latest Redis server"

NorwegianDude — Fri, 24 Jul 2026 23:39:35 +0000

While most don't, there are developers who actually do understand systems from end to end. Most developers are terrible developers, compared to the really good ones.

New comment by NorwegianDude in "It's not about physical vs. digital games, it's about ownership"

NorwegianDude — Sun, 05 Jul 2026 23:42:57 +0000

I've been running a small game dev studio for ~20 years, and the one change I think must be made, is to ban the usage of "buy" when it comes to games. Games are licensed, not bought, and that should be crystal clear to those who are paying.

Most of the games people play runs using proprietary software and/or licenses, and often on very specific hardware, with game features that makes sense for the amount of players the game has. Requiring that people should be able to play such games if the company stops running it would completely change and limit how games are developed, and in many cases require a completely different version of the game to be co-developed in case people stop playing it. It would with 100 % certainty result in slower development, fewer games, and worse games.

I do of course think that developers should try to make games playable without the company being involved, within reason. Some games that do not have licensing issues or complicated server backends as a requirement could be made available without too much work. But for things like e.g. MMORPGs it's nearly impossible. If your ever developed bigger software systems you know how many moving parts are involved, so just imagine the difficulty of making it work on consumer machines...

New comment by NorwegianDude in "Claude Sonnet 5"

NorwegianDude — Tue, 30 Jun 2026 21:23:21 +0000

A common extreme misconception is that inference is expensive and that providers are loosing a lot of money. Inference is extremely lucrative and profitable.

New comment by NorwegianDude in "U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations"

NorwegianDude — Sat, 27 Jun 2026 21:19:07 +0000

But you are talking about Europe. If the US and Europe were to cut all ties they would both face some serious consequences, and the US wouldn't be able to do anything about that, not strategically or military. The US need trust in order to function, and attacking would loose them all trust globally, making it much much more severe than anything any other nations would struggle with.

New comment by NorwegianDude in "U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations"

NorwegianDude — Sat, 27 Jun 2026 20:57:51 +0000

Hey! We Norwegians have on average more ownership in US tech companies than Americans!

Everything is going according to plan. When we have majority, we'll move the companies to Norway, but don't tell anyone, this is supposed to be a secret. /s

New comment by NorwegianDude in "U.S. allows Anthropic to release Mythos AI to ‘trusted’ US organizations"

NorwegianDude — Sat, 27 Jun 2026 20:50:38 +0000

To be fair, it would be a bigger issue for the US. No country is more economically dependant on the rest of the world than the US. The US is living on the USD, and if others stop using it the US would have to do extreme cuts on everything.

New comment by NorwegianDude in "GLM-5.2 is the new leading open weights model on Artificial Analysis"

NorwegianDude — Wed, 17 Jun 2026 12:19:26 +0000

The price, processed tokens, and output can be anything, it just depends on what GPU it is.

Nvidia GPUs are much more efficient than Apple hardware for inference(and training).

New comment by NorwegianDude in "Apple reveals new AI architecture built around Google Gemini models"

NorwegianDude — Mon, 15 Jun 2026 08:51:27 +0000

Cause I'm not trying to fool you in order to make money off you, and it should not be hard to understand that it is impossible to create a usable reply to a text someone sends you if you can't see the text.

New comment by NorwegianDude in "Under-16s to be banned from social media, Starmer announces"

NorwegianDude — Mon, 15 Jun 2026 08:25:02 +0000

The proper way to implement it is to issue digital IDs and use ZK proofs to verify the age. That way the service doesn't know anything other that the fact that you have an official digital ID and that you are at least a certain age. The ID issuer does not need to be involved in anything other than issuing the ID, making it perfect when it comes to privacy, while still fulfilling the goal of having an age limit.

If this is built on open standards, so that anyone can use it for free, it would be a big positive step forward for everyone.

New comment by NorwegianDude in "Apple Foundation Models"

NorwegianDude — Mon, 15 Jun 2026 08:10:19 +0000

A dark, but not totally unfair take: It makes it easier for Apple to take payment for the models others provide, and even allows Apple, if they want to, to use the data to build a dataset for training their own models based on how users use third party models. It's only on Apple devices this API is used, so they split up the market by not letting developers use the same system if they want things to work on iOS, locking users even more in.

New comment by NorwegianDude in "Pokémon Go Scans Trained the Navigation Tech for Military Drones"

NorwegianDude — Thu, 11 Jun 2026 12:35:57 +0000

You are creating a 3D model when you scan using Pokémon Go. Difference in lenses doesn't matter, that only matters for the scanning step.

New comment by NorwegianDude in "Apple reveals new AI architecture built around Google Gemini models"

NorwegianDude — Tue, 09 Jun 2026 01:21:50 +0000

No reasonable person relies on Apple's statements as facts. That said by Apple's laywers, and is especially true here. Apple can access the data if they want, and so can Google and Nvidia.

New comment by NorwegianDude in "Apple reveals new AI architecture built around Google Gemini models"

NorwegianDude — Tue, 09 Jun 2026 01:08:17 +0000

Exactly. That page objectively proves that Apple can't be trusted and breaks their promises.