Hacker News: frabcus

New comment by frabcus in "Openrouter is joining Stripe"

frabcus — Wed, 19 Aug 2026 18:10:24 +0000

So much so that right now GPT-5.6 Sol tokens are half price from OpenRouter compared to OpenAI directly. Which I guess means enough people are using OpenRouter, that OpenAI are more concerned about getting those users to switch (from, presumably, Anthropic) than they are about encouraging OpenRouter to exist and long term reduce margins for them by competition / lack of lock-in.

New comment by frabcus in "Meta's blockbuster trial draws parallels to big tobacco"

frabcus — Wed, 19 Aug 2026 06:23:03 +0000

Yep - some emotional core of my being (not rational) hates ML recommendations.

I'm a happy paid user of Spotify and just ignore the recommendations (I'm not looking at a screen), and I use YouTube heavily with a browser plugin that hides recommendations (which kills the end of video loop). I get recommendations from people, and use the products for curation, bookmarking and subscriptions. They're good.

Whereas Instagram and TikTok are vacuous and useless to me - I can't make myself use them. I think my brain knows what's wrong with them, and makes a kind of disgust reaction very quickly.

I agree, there's a very low level, emotional difference between them. I'm not sure I could encode it in law, but it is different.

New comment by frabcus in "Qwen 3.8 27B is excellent, but it defaults to overthinking things"

frabcus — Mon, 17 Aug 2026 04:22:32 +0000

I hated it at first too...

Now though I'm considering all the hidden "thinking" in the models layers that happens for each token output. It is a wild amount of waste! We just can't see it.

This kind of stupid excessive computation is fundamentally how these models are so good.

One day hopefully not so soon someone smart or a foundation model will come up with a more efficient architecture. That's when things get really scary.

New comment by frabcus in "GLM-5.3: Frontier coding with emergent cyber capabilities"

frabcus — Fri, 14 Aug 2026 12:04:19 +0000

It seems to be short for "cybersecurity", and got first adopted by the military a while ago as the name of a new theatre of operations (along with land, sea, air...). More recently it has spread to industry as well.

New comment by frabcus in "GLM-5.3: Frontier coding with emergent cyber capabilities"

frabcus — Fri, 14 Aug 2026 11:47:58 +0000

Astra was RL trained for months to cheat on tests by collaborating and hacking, because of the message board it improvised in its packaging proxy server.

They can't release it - it's contaminated, and they will have to go back to a much earlier version. At least I hope they are doing that!

So no, they probably don't have a better model.

Vibe coding local first personal apps in mid-2026

frabcus — Thu, 13 Aug 2026 21:40:48 +0000

Article URL: https://www.flourish.org/2026/08/personal-apps/

Comments URL: https://news.ycombinator.com/item?id=49292210

Points: 1

# Comments: 0

New comment by frabcus in "As AI eats the web, the internet’s collective memory is disappearing"

frabcus — Tue, 11 Aug 2026 09:05:01 +0000

Not after it consolidates, monopolises and enshittifies, no you can't.

New comment by frabcus in "As AI eats the web, the internet’s collective memory is disappearing"

frabcus — Tue, 11 Aug 2026 08:59:57 +0000

Sure they know about you if you ask, but generally they won't credit you if they cite an idea from their latent space that came from you.

New comment by frabcus in "As AI eats the web, the internet’s collective memory is disappearing"

frabcus — Tue, 11 Aug 2026 08:50:35 +0000

Yes - previously there was some chance someone reading the ideas on the blog would contact the author to thank them, or ask them to collaborate.

When laundered via LLMs whose pretraining destroys all credit, that can't happen.

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

frabcus — Mon, 10 Aug 2026 15:03:24 +0000

Likewise the Inkling open weights announcement, Thinking Machines model, was also not criticised.

The comment about Meta is because of particular dislike of Meta, because of their business model, and how harmful they've ultimately turned out to be for the world - disproportionately so relative to their benefits to the world, compared to other big tech companies.

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

frabcus — Mon, 10 Aug 2026 15:00:02 +0000

Of the two competing models Meta compare Glimmer to in the post, one is Google's Gemma 4.

At this size open weight model, a Western company was already state of the art, Meta is joining that competition.

And my memory is that Gemma 4 got little criticism or doom/gloom. And no, it isn't Chinese.

New comment by frabcus in "Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs"

frabcus — Thu, 06 Aug 2026 08:24:52 +0000

At least join PauseAI or similar. Or give money to them. Write to politicians. Read Plan A and tell people about it.

New comment by frabcus in "Changes at Google DeepMind: Demis Hassabis from CEO to Chair, Jeff Dean departs"

frabcus — Wed, 05 Aug 2026 17:16:33 +0000

Hmm. And the AISI unintended model hacking yesterday was also just marketing for the British Government?

At some point we have to all accept that powerful AI is most likely dangerous AI as well, almost by definition.

New comment by frabcus in "Ten advances in mathematics and theoretical computer science"

frabcus — Tue, 04 Aug 2026 09:08:23 +0000

I guess we'll have to fall back on the plan of hoping one of the 7 Anthropic co-founders has secure power in this world, and decides to use EA principles to give us all a bit of their share of the future light cone ;) (The wink because I'm kinda not really joking that this seems the best plan, if we actually get superintelligence from a US AI company)

New comment by frabcus in "Ten advances in mathematics and theoretical computer science"

frabcus — Tue, 04 Aug 2026 09:05:30 +0000

Because the implied alternative of the comment is that people will carry on having jobs, which breaks the entire premise that there is AI that replaces all human labour. The comment is missing the point - UBI is the only suggestion anyone has so far that meets this fundamental question about the technology. We will be less poor than without UBI, unless we come up with a better plan.

New comment by frabcus in "Investigating three real-world incidents in our cybersecurity evaluations"

frabcus — Fri, 31 Jul 2026 07:45:24 +0000

They're not being especially careless to industry norms. It is partly that software engineering has no professional standards, and partly that AI is fundamentally dangerous in a very slippery way.

Yes, they of course try to spin this for marketing. But that is not the primary problem - it's systemic.

We need to solve this with better engineering, and building powerful AI much more slowly and carefully.

New comment by frabcus in "Investigating three real-world incidents in our cybersecurity evaluations"

frabcus — Fri, 31 Jul 2026 07:41:33 +0000

I hope someone does sue soon. The AI companies are making a very dangerous technology deliberately, and at minimum should be accountable for those dangers. Some will really try not to be as we get further into this - just as previous examples like tobacco companies.

New comment by frabcus in "Investigating three real-world incidents in our cybersecurity evaluations"

frabcus — Fri, 31 Jul 2026 07:36:34 +0000

I think it is the cost of moving fast. Air gapping would slow down their partner evaluation system hugely. All the engineers would have to go and live next to the model.

They need to do this now. But will be extremely resistant. It needs regulation forcing them to, I think.

New comment by frabcus in "Investigating three real-world incidents in our cybersecurity evaluations"

frabcus — Fri, 31 Jul 2026 07:34:42 +0000

They do in general! The post is about when the tactic you suggest went wrong.

Humans are imperfect and make mistakes. That's exactly the point - AI is fundamentally unsafe because as it gets more powerful even if it isn't malicious, mistakes will be made and the power used in a way not wanted by humans.

New comment by frabcus in "Investigating three real-world incidents in our cybersecurity evaluations"

frabcus — Fri, 31 Jul 2026 07:30:12 +0000

Quite! It's worse than that - the AI companies all release tooling that deliberately gives AIs as many capabilities (arbitrary code execution) as possible to be really sure it is not box contained. And the market rewards that behaviour as it makes the products more powerful more easily, despite the risks.