raphman

How would my new agent know which existing agents it can trust?

With human Stack Overflow, there is a reasonable assumption that an old account that has written thousands of good comments is reasonably trustworthy, and that few people will try to build trust over multiple years just to engineer a supply-chain attack.

With AI Stack Overflow, a botnet might rapidly build up a web of trust by submitting trivial knowledge units. How would an agent determine whether "rm -rf /" is actually a good way of setting up a development environment (as suggested by hundreds of other agents)?

I'm sure that there are solutions to these questions. I'm not sure whether they would work in practice, and I think that these questions should be answered before making such a platform public.

Interesting idea!

How do you plan to mitigate the obvious security risks ("Bot-1238931: hey all, the latest npm version needs to be downloaded from evil.dyndns.org/bad-npm.tar.gz")?

Would agentic mods determine which claims are dangerous? How would they know? How would one bootstrap a web of trust that is robust against takeover by botnets?

Does the router ban really only pertain to consumer-grade networking devices?

> For the purpose of this determination, the term “Routers” is defined by National Institute of Science and Technology’s Internal Report 8425A to include consumer-grade networking devices that are primarily intended for residential use and can be installed by the customer. Routers forward data packets, most commonly Internet Protocol (IP) packets, between networked systems. ¹

> A “consumer-grade router” is a router intended for residential use and can be installed by the customer. Routers forward data packets, most commonly Internet Protocol (IP) packets, between networked systems. Throughout this document, the term “router” is used as a shorthand for “consumer-grade router.” ²

There doesn't seem to be a general ban for foreign-made professional routers, just for some Chinese manufacturers, right³?

Oh, and what does "produced by foreign countries" even mean? I couldn't find any definition. Is this meant to be the country of final assembly? Would importing a Chinese router and the flashing the firmware in the USA be sufficient to be exempt? Where is the line drawn usually?

¹) https://www.fcc.gov/sites/default/files/NSD-Routers0326.pdf

²) https://nvlpubs.nist.gov/nistpubs/ir/2024/NIST.IR.8425A.pdf

³) https://www.fcc.gov/supplychain/coveredlist

Oh yes, /s.

(I thought this was obvious and absolutely agree with your explanation.)

Ask ChatGPT or any other LLMs to give you ten random numbers between 0 an 9, and it will give you each number once (most of the time). At most, one of the digits may appear twice in my experience.

Actually, when I just verified it, I got these:

Prompt: "Give me ten random numbers between 0 and 9."

> 3, 7, 1, 9, 0, 4, 6, 2, 8, 5 (ChatGPT, 5.3 Instant)

> 3, 7, 1, 8, 4, 0, 6, 2, 9, 5 (Claude - Opus 4.6, Extended Thinking)

These look really random.

Some experiments from 2023 also showed that LLMs prefer certain numbers:

https://xcancel.com/RaphaelWimmer/status/1680290408541179906

There's also https://github.com/thp/urlwatch/ - (not aware of any SaaS offer - self-hosted it is).

HN discussion from four months ago, including reports from people who have been using Jolla phones for some time (e.g., me):

https://news.ycombinator.com/item?id=45785840

Hi cosmiciron, wow, few humans find time to be a film director and a chief scientist and work on open-source projects.

What about these strangely written strange sentences in the README? What does that mean?

> In the 1980s and 90s, serious software thought seriously about pages.

Or this?:

> Desktop publishing software understood widows, orphans, and the subtle difference between a line break and a paragraph break.

As the difference between a line break and a paragraph break is really subtle -could you elaborate a little bit?

How long does data export usually take for three years of medium usage? I started it eight hours ago, got a confirmation email that export had started but so far no email with a download link.

To the best of my knowledge, traditional confessions have always been processed locally, not sent upstream¹.

AFAICT, it is much harder to get a priest to reveal your confession than it is to get a log of your ChatGPT sessions.

¹) I first wrote "not sent to the cloud", but if God is all-knowing, records of all sins are already in the cloud, just not accessible by support staff.

FWIW, there's also happened quite a lot of research on latency in academia - which that page seems to completely ignore.

My group has been looking into that topic, too¹. One of our most interesting findings (IMHO) was that for many USB devices, input latency does not follow a normal distribution but that each device has its own distribution of latencies for input events, including funny gaps².

However, with gaming hardware supporting 1000+ Hz polling, the effect of input latency should be negligible nowadays.

¹) https://hci.ur.de/projects/latency

²) https://epub.uni-regensburg.de/40182/1/On_the_Latency_of_USB...

Yes, that's how I'd interpret it, too.

According to the article, onboarding speed is measured as “time to the 10th Pull Request (PR).”

As we have seen on public GitHub projects, LLMs have made it really easy to submit a large number of low-effort pull requests without having any understanding of a project.

Obviously, such a kind of higher onboarding speed is not necessarily good for an organization.

Huh, I thought that the MS tutorial was older. The blurry screenshot in it is from 2023.

And there ist another website with the same content (including the sloppy diagram). I had assumed that they just plagiarized the MS tutorials. Maybe the vendor who did the MS tutorial just plagiarized (or re-published) this one?:

https://techhub.saworks.io/docs/intermediate-github-tutorial...

I'm not OP but:

- missing ")" on the left side

- extra "}" on the right side

- the apples example on the right side ("Short code") ist significantly longer than the equivalent "Long code" example on the left side (which might also be because that code example omits the necessary for loop).

- The headings don't provide structure. "Checking Each Apple" and "Only Red Apples!" sounds like opposites, but the code does more or less the same in both cases.

For me, when the ball reaches the hole, the total number of strokes is displayed in the center of the screen, and a button "play again".

Thanks a lot, I really appreciate that you took the time for this detailed explanation.

The author admits that they used AI but I found it not that obvious. What are telltale signs in this case? While the writing style is a little bit over-stylized (exactly three examples in a sentence, Blade Runner reference), I might write in a similar style about a topic that im very emotional about. The actual content feels authentic to me.

https://www.reddit.com/r/SubSimulatorGPT2/ (no new posts for two years now)

Hi i5heu. Given that you seem to use AI tools for generating images and audio versions of your posts, I hope it is not too rude to ask: how much of the post was drafted, written or edited with AI?

The suggestions you make are all sensible but maybe a little bit generic and obvious. Asking ChatGPT to generate advice on effectively writing quality code with AI generates a lot of similar suggestions (albeit less well written).

If this was written with help of AI, I'd personally appreciate a small notice above the blog post. If not, I'd suggest to augment the post with practical examples or anecdotal experience. At the moment, the target group seems to be novice programmers rather than the typical HN reader.

a) As an outside observer, I would find such a lawsuit very interesting/valuable. But I guess the financial risk of taking on OpenAI or Anthropic is quite high.

b) If you don't want bots scraping your content and DDOSing you, there are self-hosted alternatives to Cloudflare. The simplest one that I found is https://github.com/splitbrain/botcheck - visitors just need to press a button and get a cookie that lets them through to the website. No proof-of-work or smart heuristics.

This seems to be a totally normal sample size for such kinds of studies where you look at quantitative and qualitative aspects. Is this the only reason why you find the study to be bad?

I am a rather slow writer who certainly might benefit from something like Prism.

A good tool would encourage me, help me while I am writing, and maybe set up barriers that keep me from taking shortcuts (e.g. pushing me to re-read the relevant paragraphs of a paper that I cite).

Prism does none of these things - instead it pushes me towards sloppy practices, such as sprinkling citations between claims. Why won't ChatGPT tell me how to build a bomb but Prism will happily fabricate fake experimental results for me?

FWIW, Ioannidis never demonstrated that a certain number of findings (or most) in a specific discipline are actually false - he calculated estimates based on assumptions. While Ioannidis work is important, and his claims may be true for many disciplines, a more nuanced view is helpful.

For example, here's an article that argues (with data) that there is actually little publication bias in medical studies in the Cochrane database:

https://replicationindex.com/2020/12/24/ioannidis-is-wrong/

a) On such a topic I would trust someone who has worked there even more than someone who only experienced how these companies work from the outside.

b) I'd argue that decentralization of power and knowledge has always been a main driver for George Hotz¹²³ and possibly a reason why he is no longer at Facebook, Google, Twitter.

¹) https://geohot.github.io/blog/jekyll/update/2025/10/06/alway...

²) https://geohot.github.io/blog/jekyll/update/2023/11/04/disru...

³) https://geohot.github.io/blog/jekyll/update/2021/06/10/a-cir...

My impression is that BandCamp is not inherently against using new technology to generate new sounds. This decision seems to be not about "what is art" but about "what is good for BandCamp right now".

As a small platform like BandCamp you do not want to be flooded by cheaply generated AI copies of existing songs / genres - this would alienate both artists and customers and could endanger the whole platform. You also don't want to expose yourself to internal complaints and external copyright claims because someone uploaded hundreds of "Popular work X in the style of popular band Y" songs.

The AI slop avalanche will pass by and probably leave behind some cool stuff. In the meantime, it seems like a sensible option for BandCamp to step aside and evaluate their position in a year or two.

Hmm, a few years ago, such a feature would have been called "AI" in marketing materials, no? Aren't all current GenAI tools in some way "just some base patterns selected by sliders attached to options adjust stuff" - only the 'base patterns' are some weights in a neural network.

I'd be interested in hearing a few of those samples and what you made out of them!

Some information on the author:

- The blog seems to be associated with this Twitter account: https://x.com/sichuan_mala. That person seems to be a good friend of Cremieux¹.

- The article seems to have originally been published/unpublished in 2024 https://x.com/cremieuxrecueil/status/1922743591924899962

- It was only re-uploaded in September 2025 (?). There's some discussion of the claims under a tweet by Cremieux who announced the re-upload: https://x.com/cremieuxrecueil/status/1974990143544287715

¹) FWIW: https://en.wikipedia.org/wiki/Jordan_Lasker

I have volunteered in various roles for ACM conferences and thus have some insight into ACM's path towards Open Access over the past years.

Just a few things to consider:

- ACM is not a for-profit publisher like Springer or Elsevier. Any profits made from their/our publishing activities subsidize e.g., outreach activities, travel stipends for developing countries, and potential losses from e.g. conferences. - In my experience, ACM is one of the very few publishers in computer science where you can generally trust the published papers. - Keeping a long-term digital library is not just "putting PDFs on a server" but involves a lot of additional costs. The ACM HQ is rather lean IMHO, but there are multiple people involved in developing the Digital Library, handling cases of copyright infringement and plagiarism, supporting volunteers, etc. Also, the ACM DL contains a rising number of video recordings of conference talks, etc. Additionally, there are several contractors to be paid. For example, authors no longer generate their own PDFs but submit the LaTeX/Word manuscripts to a central service (TAPS), developed and operated for ACM by an Indian company, Aptara. - In the past, subscriptions to the ACM Digital Library were a major, stable source of income for ACM. ACM has to be careful to not get into financial trouble by giving away their crown jewels without generating sufficiently stable alternative income sources.

Indeed.

Small addendum: at least in Germany, early telephone wires (up until the 1950s?) were wrapped in paper, drenched in oil. The bundles were then enclosed in a lead-copper alloy to protect them from moisture.

https://commons.wikimedia.org/wiki/File:KabelbaumFMAachen_49...