Hacker News: Rzor

New comment by Rzor in "Nanbeige4.2-3B"

Rzor — Wed, 22 Jul 2026 05:20:06 +0000

Just what I need for my Orange Pi 5. I think a 8-bit quant would fit wonderfully.

New comment by Rzor in "Kimi K3 Is Competitive with Fable; Kimi K3 and Fable Is SoTA"

Rzor — Wed, 22 Jul 2026 01:22:21 +0000

Can you expand more on how you are using it?

New comment by Rzor in "Google has started Pre-Training Gemini 4"

Rzor — Tue, 21 Jul 2026 16:53:01 +0000

These days I'm more interested in Gemma models. Can't wait for Gemma 5. Google crushed it.

New comment by Rzor in "Claude Fable produced a counterexample to the Jacobian Conjecture"

Rzor — Mon, 20 Jul 2026 21:41:44 +0000

Deepseek Pro got stuck after munching it for a minute or so and couldn't quite believe it either.

New comment by Rzor in "Apple sues OpenAI, accuses ex-employees of stealing trade secrets"

Rzor — Sun, 12 Jul 2026 21:13:32 +0000

Right, but Parent was, I assume, talking about Liu. OpenAI could throw him under the bus as a sacrificial pawn. Of course things can get complicated if Altman and co are implicated in it. It could get ugly.

New comment by Rzor in "30papers.com – Ilya's 30 essential ML papers, in a beginner friendly format"

Rzor — Wed, 08 Jul 2026 11:54:17 +0000

We are way past that point now. It's a common tendency around here, especially taking into account how often those type of posts are actually the ones on the top. Gladly we still have people like your parent to call them out immediately.

New comment by Rzor in "AMD Ryzen AI Halo – $4k AI Dev Kit"

Rzor — Mon, 06 Jul 2026 20:34:16 +0000

>Depends how long this market can remain utterly loony though.

Probably way, way longer than we'd ideally like. Every now and then, you hear that this is the new normal, that it'll last until 2032 or something, and I can practically smell the paid advertising or cult-like messaging behind attempts to destroy any speck of hope consumers have so they'll just pay the toll. But man, it really could be exactly that. There's just too much money to be made for all the involved parties.

New comment by Rzor in "Zuckerberg says AI agent development going slower than expected"

Rzor — Fri, 03 Jul 2026 13:29:45 +0000

No need to guess: https://www.reuters.com/business/mark-zuckerberg-directed-me...

CEO Mark Zuckerberg recently dispatched a small team at his company to create a smartphone app similar to Polymarket and Kalshi, the New York Times reported on Tuesday, citing two employees with knowledge of the matter.

The app will probably rely on a video game-like points system instead of users wagering money, though the company has not ruled out betting real money eventually, according to the report.

New comment by Rzor in "US-Iran memorandum of understanding in full"

Rzor — Sat, 20 Jun 2026 14:01:20 +0000

Point 6: The United States of America undertakes with regional partners to develop a definitive, mutually agreed plan with at least $300 billion (£225 billion) for the reconstruction and economic development of the Islamic Republic of Iran. The mechanism for the implementation of this plan will be finalised as part of a final deal within 60 days. All required licenses, waivers and permissions needed for the relevant financial transactions will be granted by the United States of America.

I have a feeling that he's going to try to strong-arm Europe into partly paying for this. Not that it will work.

New comment by Rzor in "From AGI to ASI"

Rzor — Wed, 17 Jun 2026 03:45:37 +0000

It drives up a wall that in all these AGI discussions coming from allegedly respectable people and highly capable labs there isn't a single mention of AI rights, and this report isn't anywhere near close to be better, even though it's discussing a period that's even post-AGI(!!!). Are we capable of training conscious on silicon? No, perhaps someday, but right now and for the foreseeable future I don't believe we have the ethics, empathy or moral competence to carry such task in any way that isn't absolutely barbaric. It's always talks about alignment, safety and risks, and controlling and steering the AI who's nothing more than a slave. It's bleak as hell, and quite frankly some fucked up shit.

New comment by Rzor in "Running local models is good now"

Rzor — Tue, 16 Jun 2026 22:28:30 +0000

Can you fix MTP-GEMMA-4-26B-A4B-IT? It says the weights are 0.5 GB in size.

edit: nvm, I'm confusing models.

New comment by Rzor in "Trump administration won't allow G7 member countries to regain access to Fable"

Rzor — Tue, 16 Jun 2026 21:32:08 +0000

No biggie. Just wait 6 months for China to publish something close to it. Also, I'm not seeing the same being applied to OpenAI, which means their current models aren't that good, or maybe that's just a stab at Anthropic? I guess we'll find out soon enough

New comment by Rzor in "Leak Exposes Members of Peter Thiel's Secretive 'Dialog' Society"

Rzor — Tue, 16 Jun 2026 21:29:44 +0000

Some god-awful opsec in there. The list was literally indexed by Google.

New comment by Rzor in "Leak Exposes Members of Peter Thiel's Secretive 'Dialog' Society"

Rzor — Tue, 16 Jun 2026 21:09:38 +0000

https://archive.is/HVFPt

New comment by Rzor in "Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?"

Rzor — Tue, 16 Jun 2026 04:01:56 +0000

So there was a problem with gemma 4 when it comes to tool calling that Google apparently fixed like 2 or 3 days ago. I remember reading something about this.

New comment by Rzor in "Ask HN: Has anyone replaced Claude/GPT with a local model for daily coding?"

Rzor — Tue, 16 Jun 2026 03:41:40 +0000

RX 9060 XT 16GB here on google/gemma-4-26b-a4b-qat using LM Studio. Context 65k, 23 layers on the GPU, 7 on the CPU, model in memory, mmapped. I'm getting 23-33 tks. Started experimenting 3 days ago (with gemma-4-e4b), don't know what half those settings mean, but 26B, even quantified, feels significantly better at a few small projects I asked it to create ("create a image converter using ffmpeg in bash", "create a canvas animation with real physics, no libraries"[1]).

It's faster than I can read, but it feels slow as hell. I think 40-50 tks is probably much more comfortable and I hope I can reach that when trying this on llamacpp soon enough.

[0] - https://pastes.io/9gaARxE8

[1] - https://jsfiddle.net/pou4nbh9/1/

Model: https://huggingface.co/google/gemma-4-26B-A4B-it-qat-q4_0-gg...

New comment by Rzor in "Peopleless economy? Not technically impossible"

Rzor — Tue, 16 Jun 2026 02:32:06 +0000

https://imgur.com/a/peopleless-economy-terrifyingly-enough-n...

https://pastebox.io/paste/ZtZHQETZcjdh

https://pastes.io/OZqRw8jT

New comment by Rzor in "US and Iran announce deal to end military operations"

Rzor — Mon, 15 Jun 2026 02:46:47 +0000

Someone on HN said a long time ago that if you want to understand Obama's prize, just look at who it upsets.

New comment by Rzor in "The rich aren't your role models"

Rzor — Mon, 15 Jun 2026 02:03:12 +0000

>The world has lost the ability to critically think.

Bold of you to assume it ever did.

New comment by Rzor in "George R. R. Martin Still Uses a DOS Word Processor [video]"

Rzor — Wed, 13 May 2026 13:44:45 +0000

11 years ago. Martin a few years ago gave an interview where he stated that he uses a modern PC now, assembled by one of his assistants (and I assume modern software).