Hacker News: Cynddl

New comment by Cynddl in "Discovery of a new OpenAI agent message board"

Cynddl — Fri, 04 Sep 2026 12:19:41 +0000

From the report:

> A few hours after they find the site, [the agents] start probing it for cross-site scripting (XSS) vulnerabilities.

New comment by Cynddl in "Banca Etica Suspends A/I's Account While Condemning the Sanctions Behind It"

Cynddl — Wed, 02 Sep 2026 17:11:25 +0000

> it seems fairly clear that A/I was providing services to organizations listed as Terrorist Organizations by the US, UK, and Canada

Let's take the UK as example. The organisation mentioned, Palestine Action, was indeed banned under terrorism laws. But on 13 February 2026, the UK High Court has ruled that the ban of Palestine Action under terrorism legislation is unlawful [1]. Why is that ‘fairly clear’ then?

[1] https://www.bbc.co.uk/news/live/c8x90q9nyzyt

New comment by Cynddl in "Banca Etica Suspends A/I's Account While Condemning the Sanctions Behind It"

Cynddl — Wed, 02 Sep 2026 15:58:08 +0000

> among the groups receiving those services was the PKK

Do you have any evidence for that apart from a US executive order, the same that Banca Etica condemns for political motives?

New comment by Cynddl in "Stealing Reasoning Traces from Proprietary LLM APIs"

Cynddl — Tue, 11 Aug 2026 16:01:39 +0000

> The providers did not acknowledge “any security implications arising from side channels or replay attacks.” All model providers acknowledged the receipt of our report and subsequently we were unable to launch the same attacks.

I went straight to the ‘Responsible Disclosure’ section. Not surprising, but still disappointing.

New comment by Cynddl in "Diátaxis"

Cynddl — Sat, 01 Aug 2026 22:56:52 +0000

I had to search these terms, does ADR mean ‘Architecture decision record’ and C4 the ‘C4 model’? How do you combine them in documentation?

Thomson Reuters built its own AI model that now ranks among the best

Cynddl — Fri, 31 Jul 2026 21:32:53 +0000

Article URL: https://www.thomsonreuters.com/en-us/posts/innovation/thomson-reuters-built-its-own-ai-model-that-now-ranks-among-the-worlds-best/

Comments URL: https://news.ycombinator.com/item?id=49128751

Points: 18

# Comments: 5

New comment by Cynddl in "Now is the time to give LLMs access to the ACM digital library"

Cynddl — Tue, 28 Jul 2026 23:36:04 +0000

Good question. Unfortunately, academic knowledge is widely ‘verboten’ already. Everything under paywall, researchers having to pay up to $10,000 to publish in open access in some venues, rare books unavailable even to top universities. Access to knowledge and information is increasingly difficult for everyone.

That said, what matters here is the social contract, what do I bring to society and what do we get from tech companies. For most people around the world, access to the typical leading models is out of reach. Not many on this planet can pay the subscriptions (or even API keys) that offer access to the best models. So I'm not buying the argument that tech companies are broadening access. What we're creating is a increasingly discriminatory society where the few get access to information, and the many don't.

New comment by Cynddl in "Now is the time to give LLMs access to the ACM digital library"

Cynddl — Tue, 28 Jul 2026 20:30:53 +0000

The issue here is ACM focusing on licensing. A non profit would not be able to pay ACM for access. Hence why this policy is hypocritical: it gives more power to the larger players and undermines smaller actors in the field who have fewer resources.

New comment by Cynddl in "Now is the time to give LLMs access to the ACM digital library"

Cynddl — Tue, 28 Jul 2026 20:29:12 +0000

This makes no sense whatsoever. How would a philosopher of science, or a social scientist who publish in the ACM apply for a patent? Or someone who builds software (software patent not so easy to get ;)). I have applied for patents before and I'm pretty sure my patent application has been fed to countless LLMs by now.

The issue is not who owns knowledge, it's how it benefits humanity.

New comment by Cynddl in "Now is the time to give LLMs access to the ACM digital library"

Cynddl — Tue, 28 Jul 2026 19:45:53 +0000

As a researcher with many articles in the ACM library, I have to say this is a masterclass in hypocrisy. Obviously, lawyers can decipher the terms of ACM publishing contracts and Creative Commons licences to determine if this will be acceptable or not. But ACM is not a company, it's a non-profit founded in 1947 to represent scientists.

I would be surprised if a majority of ACM members were to say yes should we ask them (but ACM is not known for such democracy). Along with book authors, we are one of the many people that provide the knowledge and expertise on which large tech firms train their models, and get nothing in return. Actually, life is getting worse for us: extra workload in universities with students' AI use, a completely broken peer review system, etc. Hence the irony of ACM thinking about licensing, and only licensing, at a time where this is the least of our priorities.

New comment by Cynddl in "AI companies are shredding rare books"

Cynddl — Mon, 27 Jul 2026 12:56:24 +0000

> For now these books are in corpuses of training data, but eventually I trust they will make their way to the rest of us.

What makes you think they will? What would be the incentives for these companies to do so?

New comment by Cynddl in "AI companies are shredding rare books"

Cynddl — Mon, 27 Jul 2026 12:54:33 +0000

The 404media article mentions notably https://nltimes.nl/2026/06/25/rare-book-dealers-fear-tech-fi... which says:

> The attachment contained 3,000 English-language titles organized by ISBN number, including books such as Distinct Element Modelling in Geomechanics by K.R. Saxena (1999); Barrett's Traditional Fairy Tales (2021), an academic study of Irish folklore; and Laser Shock Peening of Advanced Ceramics by Pratik Shukla (2018).

New comment by Cynddl in "Pseudpocalypse"

Cynddl — Thu, 16 Jul 2026 20:59:05 +0000

Entropy is unfortunately a very bad metric to estimate if these identification techniques will scale. Plugging my work here as example: https://www.nature.com/articles/s41467-024-55296-6

Hijacking Defensive Cyber AI Agents for Remote Code Execution

Cynddl — Wed, 08 Jul 2026 23:35:29 +0000

Article URL: https://ainowinstitute.org/publications/friendly-fire-exploit-brief

Comments URL: https://news.ycombinator.com/item?id=48838857

Points: 2

# Comments: 0

New comment by Cynddl in "Israel targeted Gaza children resulting in genocide, UN inquiry says"

Cynddl — Tue, 23 Jun 2026 10:50:03 +0000

If you look at the bottom of the page, you’ll find guidelines that mention which content is welcomed: “Anything that good hackers would find interesting. That includes more than hacking and startups. If you had to reduce it to a sentence, the answer might be: anything that gratifies one's intellectual curiosity.”

That said, I find this particularly of interest here given the growing attention to the use of algorithms and AI (including generative AI) for surveillance and targeting of palestinians.

Our evaluation of OpenAI's GPT-5.5 cyber capabilities

Cynddl — Mon, 04 May 2026 18:41:22 +0000

Article URL: https://www.aisi.gov.uk/blog/our-evaluation-of-openais-gpt-5-5-cyber-capabilities

Comments URL: https://news.ycombinator.com/item?id=48013028

Points: 2

# Comments: 0

New comment by Cynddl in "Making AI chatbots friendly leads to mistakes and support of conspiracy theories"

Cynddl — Wed, 29 Apr 2026 17:42:31 +0000

Hi all, co-author here! Happy to answer any questions about our work.

New comment by Cynddl in "Making AI chatbots friendly leads to mistakes and support of conspiracy theories"

Cynddl — Wed, 29 Apr 2026 15:11:12 +0000

(Title edited, was slightly too long)

Making AI chatbots friendly leads to mistakes and support of conspiracy theories

Cynddl — Wed, 29 Apr 2026 15:11:12 +0000

Article URL: https://www.theguardian.com/technology/2026/apr/29/making-ai-chatbots-more-friendly-mistakes-support-false-beliefs-conspiracy-theories-study

Comments URL: https://news.ycombinator.com/item?id=47949538

Points: 93

# Comments: 80

New comment by Cynddl in "Show HN: Utilyze – an open source GPU monitoring tool more accurate than nvtop"

Cynddl — Mon, 27 Apr 2026 19:38:18 +0000

This sounds super interesting and relevant. I run a small cluster with H100s (often research projects with vLLM) and being able to see not just usage but efficiency would be great.

I don't fully get the 100% utilisation vs. 1-10% real compute. Given you rely on telemetry from users to add new models, are you trying to predict how fast a model should be on vLLM, compared to how it runs in practice? What if users tweak some hyperparameters?