Hacker News: beacon294

New comment by beacon294 in "Hybrid Reasoning Agents: LLM agents that think in latent space, act through text"

beacon294 — Sun, 18 Oct 2026 22:53:21 +0000

Apparently we mostly reuse spatial reasoning and vision lol. I don't know that much about it, though.

New comment by beacon294 in "Knuth reward check"

beacon294 — Tue, 13 Oct 2026 12:19:20 +0000

shame does not work the same for everyone

For a view into problematic shame: https://www.amazon.com/dp/159285849X

New comment by beacon294 in "Qwen 3.8 follows GPT-5.5 Pro reasoning prefills"

beacon294 — Wed, 09 Sep 2026 19:43:02 +0000

I've read an embarrassing amount of Qwen 3.8 27b cot and it's nothing like this. I'm not refuting the OP, though, which is about continuation.

New comment by beacon294 in "The "$60 Gaming PC" – AMD BC-250 (2025)"

beacon294 — Sun, 06 Sep 2026 00:41:25 +0000

You can, it's kind of cool to have this capability on something gaming at such a low price tag. Its not optimal for your time but beats nothing by a LOT. And that model is pretty reliable.

https://unsloth.ai/docs/basics/dynamic-3.0-ggufs

New comment by beacon294 in "What we want is a hunter gatherer lifestyle with space age tools"

beacon294 — Sat, 29 Aug 2026 22:17:29 +0000

It just depends how far through the eye of the needle we already went, and how far we have to go. I don't want to go backwards. Humans are adapting, always locally, and never uniformly.

New comment by beacon294 in "Xiaomi: New CPU matches Apple cores single threaded, much faster multithreaded"

beacon294 — Tue, 25 Aug 2026 17:29:11 +0000

I won't belittle their hard work, but there's a qualifier here that they're bidding the most for the foundry time, or else nvidia would have the fastest chips.

New comment by beacon294 in "I like 'em thick: an apology to my English teachers"

beacon294 — Fri, 21 Aug 2026 11:51:38 +0000

Although it can be annoying to step into someone else's situational language/definitions/grammar, it can be a good rhetorical trick to shed the preexisting cultural context of a preexisting term such as depth.

New comment by beacon294 in "I like 'em thick: an apology to my English teachers"

beacon294 — Fri, 21 Aug 2026 11:39:33 +0000

What a good article, I really love the punchline at the end. If you skim the article, you can't even tell it's there.

That certainly increased the thickness of the article, for me.

New comment by beacon294 in "Unsloth Dynamic 3.0 GGUFs"

beacon294 — Thu, 20 Aug 2026 07:06:18 +0000

Try the llama.cpp fork by thetom. It's called turboquant after the technique

New comment by beacon294 in "GPT-5.6 Sol Pricing Cut by 50% on OpenRouter"

beacon294 — Tue, 18 Aug 2026 08:24:06 +0000

Check the remainder for hallucination for sure.

New comment by beacon294 in "Superlogical"

beacon294 — Wed, 29 Jul 2026 18:51:19 +0000

yes, dev tooling investment has always been disappointing. Mainly due to stingy buyers.

New comment by beacon294 in "The iPhone Upgrade Program is being replaced by Apple Upgrade"

beacon294 — Wed, 29 Jul 2026 01:43:22 +0000

LLMs are set to.increase hardware demands a lot. Although I don't necessarily think this is related to the program change.

I know this comment is deep in a thread, but to me it looks like they simply prefer to own the whole apple device payment cycle and not use partners.

BNPL has been a huge bubble outside the traditional credit bubble, this is the same. It doesn't seem like a big divergence excepting that it gives them a contact point and a stick with which to prod customer into a new phone sooner.

New comment by beacon294 in "Our Amish Language"

beacon294 — Tue, 14 Jul 2026 22:08:56 +0000

I know the word, thanks I think?

New comment by beacon294 in "Our Amish Language"

beacon294 — Tue, 14 Jul 2026 11:33:50 +0000

It's completely plausible the author's experience is a valid projection of the people they were surrounded by and also valid.

New comment by beacon294 in "Einstein's relativity rules chemical bonds in heavy elements, new research shows"

beacon294 — Sat, 11 Jul 2026 14:17:20 +0000

It would be the element underneath it which is synthetic. But it is interesting that all the elements in that row are soft or brittle in pre form or in some compounds.

New comment by beacon294 in "Honey, We Bought an AI Story"

beacon294 — Wed, 08 Jul 2026 03:24:11 +0000

Technicality everyone can do that in perpetuity :)

New comment by beacon294 in "GPT-5.5 Codex reasoning-token clustering may be leading to degraded performance"

beacon294 — Sun, 05 Jul 2026 09:31:37 +0000

I'm on the same journey but I bought a 3090 and put qwen 3.6 27b on it. It covers some things with better reliability. Obviously it doesn't have the breadth of a large model. If that's even a selling point for large models for coding?

New comment by beacon294 in "GPT-5.5 Codex reasoning-token clustering may be leading to degraded performance"

beacon294 — Sun, 05 Jul 2026 09:28:10 +0000

That's not a real equivalency. They are not necessarily the same (testing in production, hello!) And most importantly you do not have a local model because openai is not open!

New comment by beacon294 in "Qwen 3.6 27B is the sweet spot for local development"

beacon294 — Wed, 01 Jul 2026 05:55:34 +0000

I'll read it. It could be the quants too. Some quants I try are inexplicably bad, some seem better than official (or unsloth) quants... even what should be run of the mill gguf quantization.

New comment by beacon294 in "Qwen 3.6 27B is the sweet spot for local development"

beacon294 — Tue, 30 Jun 2026 20:29:51 +0000

I find that fp8 cache can be pretty bad in vllm but works fine in llama.cpp. I don't know why, but I plan to review the implementations.