How 'Artificial Intelligence' Is Slowly Drowning In Its Own Droppings

© 2026 Peter N. M. Hansteen

A website designed to appeal to humans' vanity serves to reveal how the large language models degrade by ingesting each others' unverified output

In early July 2026, I was made aware of the website intheweights.com via a fediverse post. The idea is simple, and appeals to human vanity: You type in your name, and the site checks the various large language models for any data it can find on you in the models.

---

Note: This piece is also available without trackers but classic formatting only here.

---

The post I responded to appears to be no longer available, but I checked out my own results and posted the results with screenshots:

Post by @pitrh@mastodon.social

View on Mastodon

If for some reason the embedded post is not readable, I also saved screenshots of the site output.

First up, the main result, with what seems to be the highest confidence score:

Initial highest scored intheweights.com assessment based on my name July 2 2026:
PETER N. M. HANSTEEN
OPENBSD PF FIREWALL AUTHOR
> 608 STRENGTH • TOP 6%
GPT-5.5 SAYS
NORWEGIAN TECHNOLOGIST AND AUTHOR KNOWN FOR OPENBSD, PF FIREWALL ADVOCACY, AND THE BOOK THE BOOK OF PF•"

Fair enough, I advocate OpenBSD but did not write any part of the operating system, I wrote The Book of PF, and I work in the tech industry.

I addition, there were two entries that I was pleased to see the site labeled as likely hallucinations:

Initial intheweights.com assessment 1 of 2 of likely hallucinations based on my name July 2 2026:
PETER N. M. HANSTEEN
NORWEGIAN SOLAR PHYSICIST
52 STRENGTH•
GROK 4.20 SAYS
NORWEGIAN SOLAR PHYSICIST SPECIALIZING IN THE SUN'S CHROMOSPHERE, TRANSITION REGION, AND CORONAL HEATING. "

That is definitely not me. For context, according to the Name Search service on Statistics Norway's (Statistisk Sentralbyrå) website, we are only 140 persons who have Hansteen as their surname, and Fewer than four people or no one has the name Peter Nicolai Mathias Hansteen, and to my knowledge I do not share first names with any living relative.

I am no longer active on the site that used to be called Twitter, which is where I assume Grok finds a significant subset of the data it ingests. But while there, since the late noughties until shortly after the site was sold, I may very well have posted about global warming with links to actual studies that involve solar phenomena.

---

Update 2026-08-02: Prompted by something else entirely, I checked Google Scholar (I refuse to give a direct link), which seems to confuse me with one VH Hansteen, who does publish in that field, with several publications where one H Peter is a co-author.

I am sure somebody will come up with a snappy and snarky term for that phenomenon.

---

The next entry I found a bit more bizarre, to wit:

Initial intheweights.com assessment 2 of 2 of likely hallucinations based on my name July 2 2026:
PETER NM. HANSTEEN NORWEGIAN GEOPHYSICIST

46 STRENGTH <
QWEN3 SB SAYS
A NORWEGIAN GEOLOGIST AND ACADEMIC, KNOWN FOR CONTRIBUTIONS TO GEOPHYSICS AND EARTH SCIENCES.

That one is a bit strange. While I do have a living close relative who is a geologist with publications in academic journals, this entry seems to have taken most of its data from the biography of Christopher Hansteen. We are related, but as can seen from the Wikpedia article, separated by several generations.

But at least, the site considered both of those to be likely hallucinations.

But give it a couple of weeks

For whatever reason (er, actually my laziness), the intheweights.com tab stayed open in my browser, and I went back to check the results on July 21st.

The first entry was much the same,

Highest scored intheweights.com assessment based on my name July 21 2026:
PETER N. M. HANSTEEN
OPENBSD PF fIREWALL AUTHOR
731 STRENGTH • TOP 4%
€ GPT-5. 5 SAYS
NORWEGIAN TECHNOLOGIST AND AUTHOR KNOWN FOR OPENBSD, PF FIREWALL ADVOCACY, AND THE BOOK THE BOOK OF PF.

Oddly enough, the confidence had risen a bit, and I was now in the top four percent. I still did not write that operating system, I only use it and like it enough that I will likely talk about it to strangers.

The second had apparently been moved up the ranks and was no longer considered a likely hallucination:

Second highest intheweights.com assessment based on my name July 21 2026:
PETER N. M. HANSTEEN
NORWEGIAN SOLAR PHYSICIST
43 STRENGTH • TOP 95%
GROK 4.20 SAYS
NORWEGIAN SOLAR PHYSICIST SPECIALIZING IN THE SUN'S CHROMOSPHERE, TRANSITION REGION, AND CORONAL HEATING.

So what happened here is that Grok had grown more confident in its most likely imaginary physicist. Another close but no longer living relative was a physicist, but the description does not map well to their actual published work.

Then thankfully, the last item was still considered a likely hallucination:

Third ranked intheweights.com assessment based, rated likely hallucination, on my name July 21 2026:
PETER N. M. HANSTEEN
NORWEGIAN GEOPHYSICIST
38 STRENGTH <
QWEN3 SB SAYS
A NORWEGIAN GEOLOGIST AND ACADEMIC, KNOWN FOR CONTRIBUTIONS TO GEOPHYSICS AND EARTH SCIENCES.

So the conclusion here seems to be that while the confidence in the largely fact based first item had risen overall, so had the rating for one of the likely hallucinations.

The explanation here is likely that for my material, it is very much available online, some of it under copyright but free, some under copyright but without technologically enforced restrictions, and some is actually under a free license for anyone to use as they wish.

For my relatives who are actual academics, both the living and the deceased, more than likely they have published more material than I have in total, but a large part of that material either predates the Internet or is hidden behind paywalls. Then again my relative the geologist insists that the bulk of their material is indeed freely available online with sufficient information to separate their work from mine.

So there we have it: The large language models can only ingest what is placed within reach, and they work in parallel to generate material based on what their inner, opaque, and possibly quite tangled, logic considers probable. Then it seems that the output generated soon after is fed back to the large language models as training data for themselves.

This means that more likely than not, more material will be attributed to the more visible person in the training data, based on their version of probability, than is actually appropriate. We saw an early version of this in Those Titles the AI Bot Thought I Had Written (also here).

With the large language models seeming to grow more confident of the actual existence of their own fictional creations, how likely is it that they will soon sink out of sight, buried in layer upon layer of unverified output from themselves and their peers?

My best hope is that we will be able to clean the Internet of their nonsense before it is too late.

---

How 'Artificial Intelligence' Is Slowly Drowning In Its Own Droppings is © 2026 Peter N. M. Hansteen (published 2026-07-22)

---

You might also be interested in reading selected pieces via That Grumpy BSD Guy: A Short Reading List (also here).

Those Titles the AI Bot Thought I Had Written

© 2024 Peter N. M. Hansteen (2024-12-06)

Beware of robots generating your references. They could very well take it upon themselves to lie on your behalf. Testing in 2023, I observed a whopping 25% truth in the result. What is the truth rating in 2024?

Note: This is a field note that was originally intended as a fediverse (mastodon) post, but the text would have overflowed into a thread and would then have become quite cumbersome.

Back in the early days of generative Artificial Intelligence revolution (read: late 2023) I had heard enough of the hype that I did a few minimal experiments (see my earilier notes ChatGPT opines on IPv6 procastination, waxes lyrical over OpenBSD (also prettified, tracked) and I asked ChatGPT to write a pf.conf to spec, 2023-06-07 version (also prettified, tracked)) that had me conclude that for my own purposes at least, the technology as it appeared at the time was unlikely to improve productivity for myself or any of my immediate colleagues.

The TL;DR would be that since the models will only become somewhat useful if they are trained on a sufficient volume of material within the topic of interest, clearly a lot of work needed to be done. Basically, the gaps in the training data seemed to fit essentially most of the field I deal with professionally, beyond of course a few trivial generalities. And those were cracks that could only be papered over at the cost of pulping several good-sized forests.

---

Note: This piece is also available without trackers but classic formatting only here.

---

Not easily discouraged, I thought maybe, it just might be, that the data the language model trainers had sucked in from public sources on the Internet would contain enough factual information that the system would be able to cough up a reasonable bibliography of what I had written.

So I opened another session with the same robot and typed

---

what are the most read articles by Peter N. M. Hansteen?

---

After a brief pause, the cursor started moving and presented its result:

---

As an AI language model, I don't have real-time access to up-to-date information or specific statistics on article readership. Additionally, my training only goes up until September 2021, so I may not be aware of the most recent articles published by Peter N. M. Hansteen. However, I can provide you with some notable articles that Peter N. M. Hansteen has written up until my knowledge cutoff date.

Peter N. M. Hansteen is a well-known author and has written extensively on the topic of networking and computer security, particularly focused on OpenBSD and firewalls. Some of his notable articles include:

"The Book of PF: A No-Nonsense Guide to the OpenBSD Firewall" - This is a comprehensive guide that provides practical information on setting up and configuring firewalls using OpenBSD's PF (Packet Filter) system. It covers topics such as filtering network traffic, NAT (Network Address Translation), and quality of service (QoS).

"Building Firewalls with OpenBSD and PF" - This article discusses the concepts and techniques involved in building firewalls using OpenBSD and its PF system. It provides practical examples and guidance on firewall design and implementation.

"Securing the Edge: A Practical Introduction to OpenBSD Packet Filtering" - This article focuses on the fundamentals of packet filtering using OpenBSD's PF system. It covers topics such as rule syntax, filtering criteria, and advanced filtering techniques.

"So You Want to Be a Firewall Wizard? Part 1" and "Part 2" - These two-part articles provide an introduction to firewall concepts and techniques, with a focus on OpenBSD and PF. They cover topics such as rule sets, network address translation, and handling complex firewall configurations.

Please note that the popularity of articles can vary over time, and it's always best to refer to the author's official website, blog, or other reliable sources for the most up-to-date information on their publications and readership.

---

(Preserved as text, cut and pasted from the website, here)

My initial reaction when I saw the result was a sound you would associate with something quite unpleasant happening.

I'll take one paragraph at the time.

Though it was probably intended as a disclaimer, the claims in the first paragraph there are, as we shall see, only 25% true, if even that.

The second paragraph has a description of me that is on par for what would turn up in a not terribly specific search using your favorite Internet search engine.

Just how well known an author and what consitutes extensive writing is open for discussion, but at least the paragraph as a whole is not counter-factual.

I did write The Book of PF. The third edition of the book would have been available via that link and a few others at the September 2021 cutoff date, and the description is a valid, if somewhat short of exhaustive, description of the book when it comes to subjects covered.

However, the three titles that follow were not written by me.

I think I have read articles matching those titles at some point, but the actual authors of those pieces would not be terribly pleased to have them attributed to somebody else. I would assume their publishers would like a word there too.

I have mentioned this session a few times to colleagues and friends over refreshments, but I have not really give it much thought since.

But the chatter about how the AI bot herders are heroically fighting their charges' tendency to hallucinate seems to never come to an end.

So my question to you is: Has the situation actually improved since then?

If you have any useful experience to share or other input on the subject, please follow up with a comment in the thread where you found the reference to this note.

A reasonable followup question would be,

Are the efforts that go into (what is currently labeled as) Artificial Intelligence starting to pay off?

followed by

or are we for now better served by having the abbreviation "AI" (perhaps more reasonably) expand to "Artificial Inanity"?

Come to think of it, unless the various bots that scrape the net actually obey my robots.txt directives, publishing even this text here will lead to further confusion over who actually wrote those titles. This is bound to turn up in the AI slop presented to some poor soul looking for valid information.

Again, if you have any useful insigths or experiences you want to share on the subject, please follow up with a comment where you found the link to this piece, or post to this fediverse thread (raw link https://mastodon.social/@pitrh/113612679976829404).

I look forward to reading your comments.

---

Those Titles the AI Bot Thought I Had Written is © 2024 Peter N. M. Hansteen (published 2024-12-06)

---

You might also be interested in reading selected pieces via That Grumpy BSD Guy: A Short Reading List (also here).

At EuroBSDcon 2025, there will be a Network Management with the OpenBSD Packet Filter Toolset session, a full day tutorial starting at 2025-09-25 10:30 CET. You can register for the conference and tutorial by following the links from the conference Registration and Prices page.

Separately, pre-orders of The Book of PF, 4th edition are now open. For a little background, see the blog post Yes, The Book of PF, 4th Edition Is Coming Soon. The latest information I have is that physical copies will be shipping by the end of January 2026

---