Hacker News: cubefox

Pfizer, Valneva Lyme vaccine starts review in Europe

cubefox — Mon, 17 Aug 2026 22:26:29 +0000

Article URL: https://pharmaphorum.com/news/pfizer-valneva-lyme-vaccine-starts-review-europe

Comments URL: https://news.ycombinator.com/item?id=49338530

Points: 1

# Comments: 0

New comment by cubefox in "Anthropic's 'watermark' text adulteration in Claude is a perversion of writing"

cubefox — Mon, 17 Aug 2026 14:02:24 +0000

> How would that work? Claude appending " written by AI" to each of its messages? That would both be impractical and useless.

New comment by cubefox in "Anthropic's 'watermark' text adulteration in Claude is a perversion of writing"

cubefox — Mon, 17 Aug 2026 08:24:35 +0000

We were talking about text.

New comment by cubefox in "A SAT Attack on Tarski's High School Algebra Problem"

cubefox — Mon, 17 Aug 2026 08:22:10 +0000

Usually when people say "not finitely axiomatizable" they just mean first-order logic. For example, Peano arithmetic is finitely axiomatizable using second-order logic, which allows quantification of predicate variables, which is required for the axiom of induction.

New comment by cubefox in "Anthropic's 'watermark' text adulteration in Claude is a perversion of writing"

cubefox — Mon, 17 Aug 2026 00:24:48 +0000

> The EU regulation, from my understanding, requires AI content to be labeled for human viewers.

How would that work? Claude appending " written by AI" to each of its messages? That would both be impractical and useless.

New comment by cubefox in "Anthropic's 'watermark' text adulteration in Claude is a perversion of writing"

cubefox — Mon, 17 Aug 2026 00:16:14 +0000

This comment disagrees with you: https://news.ycombinator.com/item?id=49324387

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

cubefox — Sun, 16 Aug 2026 23:44:08 +0000

This is just wrong. There is a huge difference between a book that doesn't explicitly tell you how to create a deadly virus, and an infinitely patient future LLM that leads you through every step and every unexpected problem and every question you might have.

New comment by cubefox in "A SAT Attack on Tarski's High School Algebra Problem"

cubefox — Sun, 16 Aug 2026 21:44:50 +0000

Not finitely axiomatizable using first-order logic, I assume.

New comment by cubefox in "Research papers using "kidney disappointment" instead of "kidney failure""

cubefox — Sun, 16 Aug 2026 18:39:39 +0000

By the way, Google Translate is really good nowadays. It would certainly not make such mistakes anymore. In my experience, for long text it is even outperforming conventional LLMs in terms of faithfulness: Some LLMs tend to produce overly loose translations on long prose. Also, sometimes they just refuse to translate text they deem offensive.

New comment by cubefox in "GPS and the Lost Art of Getting Lost"

cubefox — Sun, 16 Aug 2026 18:29:52 +0000

The park orienteering course at the end reminds me of a similar development in game design: Even before most people had GPS phones in real life, many video games already had a form of virtual GPS: A map with a marker of where the player is in the virtual world.

Only much later did various games start to abandon that marker (possibly even a map in general), because it made navigation too easy: it disincentivized exploration, orientation, things like figuring out vague verbal directions, and path finding in general.

New comment by cubefox in "GPS and the Lost Art of Getting Lost"

cubefox — Sun, 16 Aug 2026 17:51:18 +0000

In the past, people were also heavily relying on road signage or stopping and asking locals, especially when not carrying a map.

New comment by cubefox in "GPS and the Lost Art of Getting Lost"

cubefox — Sun, 16 Aug 2026 17:46:58 +0000

I'm pretty sure there is sufficient general research which says that if you don't train a skill, it gets weaker.

New comment by cubefox in "Claude: System Prompts"

cubefox — Sun, 16 Aug 2026 17:36:16 +0000

Which seems to contradict the usual consensus that purely linear architectures are not sufficiently capable and unsuited for frontier models.

New comment by cubefox in "Claude: System Prompts"

cubefox — Sun, 16 Aug 2026 14:35:03 +0000

Unless they are using a linear architecture, the compute cost still scales O(n²) for n tokens, and nemory cost scales O(n).

New comment by cubefox in "Claude: System Prompts"

cubefox — Sun, 16 Aug 2026 14:30:29 +0000

I would expect this only to be true for linear architectures like Mamba or Gated DeltaNet. Transformers and hybrid architectures do not have constant compute cost per token.

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

cubefox — Sun, 16 Aug 2026 07:20:01 +0000

> Data and content related to "biological weapons" already exist on the internet, in books, etc. The real issue is access to facilities and tools.

No, I think tools are easy to come by (unlike in nuclear research), the real issue is the know-how to create biological weapons, which you can't easily get out of books, but much more easily out of an amoral LLM.

New comment by cubefox in "At-home test for infected ticks could improve Lyme Disease diagnosis"

cubefox — Sat, 15 Aug 2026 16:16:22 +0000

There is potentially a new Lyme vaccine coming out in about a year. The review process in Europe just started.

https://pharmaphorum.com/news/pfizer-valneva-lyme-vaccine-st...

New comment by cubefox in "Phones should have a 'guest lock' feature"

cubefox — Fri, 14 Aug 2026 09:45:22 +0000

> To give a couple of examples, using this would be a bit like locking the glove box in front of traffic cops, or going around the house locking rooms when you have a guest.

That's a good point. Doing this basically says "I distrust you", which you probably don't want to say. (At least I don't.)

New comment by cubefox in "Phones should have a 'guest lock' feature"

cubefox — Fri, 14 Aug 2026 09:42:31 +0000

> Due to the intent system on Android allowing a pinned application to open another one (for example by sharing something),

I tested it and it blocks opening other apps when pinned. Maybe blocking app opening doesn't work for apps using some older system APIs.

New comment by cubefox in "Phones should have a 'guest lock' feature"

cubefox — Fri, 14 Aug 2026 09:39:34 +0000

I don't understand why this isn't activated by default. It just adds an extra "pin" option when you tap the app icon.