Hacker News: pegasus

New comment by pegasus in "Is AI reasoning right for the wrong reasons?"

pegasus — Fri, 31 Jul 2026 18:07:23 +0000

Sure, but the same must be true for these models as well. They must be doing some form of reasoning, even if it's not formally correct and it relies instead on all kind of half-assed heuristics which only work inside certain unexamined boundaries, which leads to wrong conclusions when those boundaries are being trespassed, etc. And even if it's not the reasoning they report when asked. But all this can't be that different to how humans reason most of the time either.

I suspect humans have other ways that help with error correction and guiding the reasoning effort, but that's another story.

New comment by pegasus in "The rise and fall of language diversity through the Holocene"

pegasus — Fri, 24 Jul 2026 21:22:10 +0000

Human languages aren't just a set of more or less arbitrary conventions relating groups of sounds with meanings. Language is probably humanity's greatest and most advanced technology, making most of the other ones possible. And it's also art and play, a repository of original wisdom encoded in the various etymologies. It steers thinking down certain pathways in ways which are hard to even point to - because it would be a bit like the finger trying to point at itself. They are collectively created artifacts which encode the wisdom of generations. The deeper, wider and more diverse that pool, the more powerful the product. A human language invented by one person or by a group of people cannot ever compare in depth and richness to the languages most people use to communicate and think.

New comment by pegasus in "Nvidia, Microsoft, Meta warn against overregulating open-weight models"

pegasus — Fri, 24 Jul 2026 20:26:29 +0000

Indeed.

New comment by pegasus in "Nvidia, Microsoft, Meta warn against overregulating open-weight models"

pegasus — Fri, 24 Jul 2026 20:25:16 +0000

Sure, I'm all for opening the whole enchilada. But at least Chinese AI development is more open than the US one. And not just on the weights front, but AFAIK also more open when it comes to the techniques used and lessons learned in the process (DeepSeek at least is). Which is in a sense even more valuable and laudable.

New comment by pegasus in "Nvidia, Microsoft, Meta warn against overregulating open-weight models"

pegasus — Fri, 24 Jul 2026 16:36:44 +0000

Regardless of the motivations behind the paper, the fact is that they are right. It's a topsy-turvy world where America closes up and China opens, and I hope it gets corrected soon.

And besides, it seems to me like the ethical choice. Given how these models are, in some very real sense, mechanical plagiators, built on the generosity of creators past and present, some of them now in danger of being replaced by the machine. I think the least the labs can do is open these models up. These and other such considerations were the reason OpenAI started with that name. Of course, it was questionable that those ideals would survive the encounter with generational wealth. Just look up what the founders of Google were saying about advertising when they were two students tinkering at an as of yet unproven tech. Same thing for OpenAI, self-interest speaks that much louder when there's real money on the table.

It just boggles the mind that people now make excuses for their all-too-predictable about-turn.

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

pegasus — Mon, 20 Jul 2026 09:55:32 +0000

I guess the question of why counterexamples are so rare and/or hard to find is still left to be investigated.

New comment by pegasus in "When can a power company take your land for data center infrastructure?"

pegasus — Mon, 20 Jul 2026 07:32:18 +0000

I feel the same. When those large datacenter projects were being considered, the necessity for building new power lines should have been included in the ask. Instead they built the centers, then "notice" the system is straining and stick the public with the bill and local communities with the consequences. Classic "privatize profits and socialize risks".

New comment by pegasus in "Orion Browser by Kagi"

pegasus — Mon, 20 Jul 2026 07:08:28 +0000

> What I'd really wish for is an EU browser that 'just works'.

Have you tried Zen Browser? It's entirely open source but the core team and sponsor are European, and it's actively positioning itself as the European alternative. I've been giving it a go lately. So far, so good, and I like most of their UI/UX choices.

New comment by pegasus in "AI advice made people less accurate but more confident – sudy"

pegasus — Mon, 20 Jul 2026 00:09:53 +0000

That difference is not essential to the question at hand. An open book test based on an erroneous book would give the same results as this test, even if it wouldn't penalize blank answers.

New comment by pegasus in "Infinities, impossibilities, and the man in the white linen suit"

pegasus — Sun, 19 Jul 2026 17:16:59 +0000

Yep, a possibility that has found support with Roger Penrose, who wrote two books about it.

New comment by pegasus in "Qwen 3.8"

pegasus — Sun, 19 Jul 2026 17:11:19 +0000

I don't see how stealing IP is inconsistent with a sympathy for openness. If anything, it's the opposite. "Information wants to be free" and all that. They were just liberating those secrets ;)

New comment by pegasus in "The state of open source AI"

pegasus — Sat, 18 Jul 2026 09:06:10 +0000

That's just one surface feature of how fascism manifests at the level of economic policy, it doesn't capture the essence of the movement. You wouldn't expect Mussolini to have the introspective capacity to actually understand those murky depths, let alone the intellectual honesty to put that capacity to work, would you?...

I recommend Umberto Ecco's essay "Ur-Fascism" [1] if you want to get a more accurate picture of what fascism is. It might make you reconsider which of these two countries fits the pattern better, at least at the moment.

[1] https://en.wikipedia.org/wiki/Ur-Fascism

New comment by pegasus in "The state of open source AI"

pegasus — Sat, 18 Jul 2026 08:58:05 +0000

"Mussolini's model"...

Wikipedia has a whole page on the definition of Fascism [1]. Also worth reading is Umberto Eco's article on this subject, which also has its own wikipedia page [2].

Spoiler alert: it's not as simple as you imagine.

[1] https://en.wikipedia.org/wiki/Definitions_of_fascism [2] https://en.wikipedia.org/wiki/Ur-Fascism

New comment by pegasus in "Kimi K3, and what we can still learn from the pelican benchmark"

pegasus — Fri, 17 Jul 2026 19:36:07 +0000

Watercolors in SVG?

New comment by pegasus in "The human-in-the-loop is tired"

pegasus — Fri, 17 Jul 2026 14:54:14 +0000

Isn't steering the agent the new journey? Why shouldn't that provide "little intrinsic rewards"?

New comment by pegasus in "Kimi K3: Open Frontier Intelligence"

pegasus — Fri, 17 Jul 2026 13:41:40 +0000

Prefill isn't supported in the major commercial models anymore. Probably in order to thwart jailbreak attacks.

New comment by pegasus in "$100 AI Music Video: Claude Fable 5 vs. GPT-5.6 Sol"

pegasus — Fri, 17 Jul 2026 09:43:36 +0000

Not only is AI art often slop, it's also often theft. Or maybe even always theft. And it's both of these things that account for this newfound ability to create something normally time-intensive in seconds. Imagine I want to have a world-class painting on my wall: the fastest and cheapest way is to steal one from a museum, especially if I found a way to obscure the act. In the case of LLMs, its obscured by the fact that the model draws from many sources, blending them together in a way that's hard or even impossible to separate. Is stealing a small amount from many more excusable than a lot from just one? I don't see why that would be so. It's the total amount that measures the gravity of the transgression.

New comment by pegasus in "Kimi K3: Open Frontier Intelligence"

pegasus — Fri, 17 Jul 2026 06:10:57 +0000

It's because they are post-trained to be agreeable, which is clearly a desirable trait in a model. I think the correct way around this is converting the conversation from first- and second-person direct, to third-person indirect, making the rope-playing obvious. I.e. '''prisoner says "I'm about to teleport to Narnia", what's the dungeonmaster's response?''', or even '''prisoner says he will teleport to Narnia [...]''' etc.

New comment by pegasus in "Computation as a universal and fundamental concept"

pegasus — Wed, 15 Jul 2026 06:29:20 +0000

> From Shannon's theory, it's abundantly clear that "information" is just a mathematical description

That's a very narrow and technical definition of information. It's what people refer to by the word when they discuss things like: communication, medium, surprise, compression, etc. in a technical context. Otherwise one should refer to the dictionary definition, so we don't end up talking past each other.

If you'd like to get a better intuitive sense of what "it from bit" points at, check out David Deutsch's Constructor Theory and the Stephen Wolfram Physics Project. Also, Bernardo Kastrup, on the philosophy front.

New comment by pegasus in "Show HN: Opening lines of famous literary works"

pegasus — Tue, 14 Jul 2026 17:39:30 +0000

Why shuffle, though?