A Song for a New Day ... & Last Call

"A Song for a New Day", by Sarah Pinkster, 2019, 384 pages, 117k words.

Wow, talk about an eerily prescient book. I thought this was new, I must have brought it for cheap on BookBub, it came out September, 2019. A female rock-n-roller is just breaking through, when terrorist attacks & a pandemic cause a shutdown of all public gatherings.

Fast forward a few years, another female becomes a talent scout for Big Music. She looks for talent in illegal underground live performance venues. Shocker, Big Music, who records 3d performances & restages them in virtualities, is not looking out for the best interests of anyone but themselves.

Not too tense, but, again, shockingly prescient of the COVID shutdowns. But interesting, defying the shutdowns in the book has a leftist feel, where during the real pandemic, it became a rightist shibboleth.

A fun & easy read, check it out. Hunh, I just noticed I flagged this book as Politics rather than Science Fiction.

---

On an impulse, I reread 1 of my fav books: "Last Call", by Tim Powers, 1992, 560 pages, 203k words.

A fisher king story, with tarot cards & weird theories of luck. What's not to like? It was longer than I expected.

I was going to power through & read books 2 & 3 of the trilogy, "Expiration Date" & "Earthquake Weather".

Kobo showed the books but wouldn't let me order them. I opened a ticket with support, a few days later, they told me it was fixed. Nope.

A few days after that, they told me it was a content problem, & they were turning it over to the content team.

A few days after that, the books were gone from Kobo. Sorry, they no longer exist.

During this whole experience, I am of course thinking, "Oh fuck, it's a goddam AI.". As the Bullshit Apocalypse proceeds, I fear we will increasingly be having to live with systems that work only statistically - they work a lot of the time, but for some other %age of the time, they fail miserably. We will live in a world of completely unreliable systems. Frustration upon frustration.

But, the techbros will get big $$$! &, all the glorious slop! I'm still saying, 2-3 years until the concensus is, "Can we get a do-over on this AI thing?"

"Bullshit All The Way Down" Recurses???

Wow, this 1st link is clearly "Bullshit All The Way Down". But, fractally? Recursively?

&, note, this was a Conde Nast creme de la creme publication. From the most excellent 404 Media:

Ars Technica Pulls Article With AI Fabricated Quotes About AI Generated Article

A story about an AI generated article contained fabricated, AI generated quotes.

---

Not 1 of our normal sources, Matthew Stoller's "Big" blog, on monolopy, antitrust, & inequality. Stoller tends to run long ...

Monopoly Round-Up: Why the "AI God" Narrative is Actually a Corporate Power Grab

The AI cultists went into overdrive this week, Meta got the creepiest patent imaginable, and AOC attacked monopolies to burnish her foreign policy chops.

---

Internet sage Doc Searls has been drinking the AI koolaid for quite a while now. He is not an ooh-ahh type. But this piece is over the top. I commented 2x in his blog:

This is a great & comprehensive piece. Thanks.

But, Doc has drunk the AI koolaid. The lists, with dozens of alternatives, of names for various new phenomenon, are overwhelming enough that I would be tempted to characterize them as “AI slop”.

AI;DR.

Life is to short to waste it sorting through “AI”-generated word salad.

Note, Doc is ~4 yrs older than me.

And

My god, can anyone read this article & have any doubt that the Bullshit Apocalypse is real & ongoing? 2 lists of ~40 bullshit names for bullshit AI phenomenon? What purpose does that serve?

Game over man, game over. We have already accumulated decades of AI slop, soon it will be centuries.

But. I’m sure, LLM-based agents will take care of cleaning it all up, chop-chop!

This is a really heavy-duty piece. It was posted Monday, Feb 16. I thought he later posted something to the effect that he was going to split this up into 4 or more pieces, but I can't find it. He is asking some fundamental questions here. If you don't subscribe to Doc Searls, you should.

Where Are We?

There are four giant problems to pull apart here:

1. How Google is deprecating the Web.
2. What to call the vast everything that Big AI now uses to give us answers and pitch us shit.
3. How AI dependence is changing us from homo sapiens to homo something else.
4. How we can stay human in the midst of it all.

---

Here's the home page/directory for my posts on Bullshit. This is post #96.

Pretenders to the Throne of God

"Pretenders to the Throne of God", by Adrian Tchaikovsky, 2026, 512 pages, 172k words. The 5th story in The Tyrant Philosophers series. Looking to be the penultimate volume.

This has been an entertaining series. I luv the covers.

The Palleseen campaign to bring atheistic perfection to the world - a world which actually has real gods & demons - encounters serious difficulties after the Pyrrhic victory of their last war.

Characters from earlier novels in the series are reintroduced, to good effect. The next novel is to be the last, & I think he has the characters on the board to pull off a most satisfying conclusion.

AI Ruins Everything

OK, I know there are many applications for which LLMs, under human guidance, can do great work.

But, I'm gonna say, for every 1 of those, there are 10 applications where LLMs & LLM-based agents are going to be turned loose to produce slop & generally screw things up.

I was so disappointed when oracle sites just became betting. I'd like a pool on "when the concensus opinion on LLMs is, can we get a do-over?"

I'm estimating 2-3 years. The techbros will probably have cashed out enough by then. The hype machine will be completely sputtering.

Meanwhile, how long to get rid of all the slop? Decades? I just today read something about using LLMs to check LLMs, why am I skeptical that this would work?

---

I still need a phrase or acronym for "LLMs screw up yet another system". LSUYAS? From the ever-excellent 404 Media:

Pinterest Is Drowning in a Sea of AI Slop and Auto-Moderation

Users are exhausted fighting AI moderation, AI-generated art, and AI-first features.

---

We've seen this before. Who cares about democracy if techbros are making big $$$? Capitalism, FTL. From The Conversation:

Swarms of AI bots can sway people’s beliefs – threatening democracy

---

Goddamit, I have only 1 issue that I really care about emotionally. That is: protect children. The goal of every government should be to maximize the outcome of EVERY child. How can we go wrong if that is our goal?

Grrr. This next makes me angry. From 404 Media:

'Students Are Being Treated Like Guinea Pigs:' Inside an AI-Powered Private School

Leaked documents reveal the inner workings of Alpha School, which both the press and the Trump administration have applauded. The documents show Alpha School's AI is generating faulty lessons that sometimes do "more harm than good."

---

Here's the home page/directory for my posts on Bullshit. This is post #95.

The Regicide Report

"The Regicide Report", by Charles Stross, 336 pages, 109k words, 2026. Book 14 of The Laundry Files.

After 22 years, Charlie Stross has wound up The Laundry Files series. Over the years, a time or 2 I wondered if it was time for Charlie to hang it up on the series. The prior novel was 1 of those times. So I was a little apprehensive, & actually delayed starting this one.

I should not have worried. I think returning to the original characters, Bob & his wife Mo, was the right thing to do. They are very complicated characters. The plotting is great. I definitely LOLed when the British royal family was described as "the world's most expensive reality TV show" - I think I have made that observation in the past.

So, thanks, Charlie, for finishing the series in such a smashing fashion.

Charlie has some new stuff cooking that I am looking forward to. But he's had health issues, & is reaching the age where he will start becoming bionic. So patience will be the watchword.

Semantic Ablation

My wife & I vacationed near Deshaies, Guadeloupe Feb 11-18. I took a vacation from my RSS feeds & SubStack.

Since we returned, I've drained all my RSS folders except for Economics. There are 16 Bullshit Apocalypse related posts.

At the start of last year, I switched (SFF) book reviews from being a few books to being 1 book/post. I like that better.

I thought about doing that with the Bullshit Apocalypse posts, but, this is just too many. So I think I'll do these mostly 3 at a time.

---

I will, of course, immediately ignore that algorithm & cover only 1 article in this post. I think this article is (by far) the most interesting & insightful of the 16 I've got queued up. By Claudio Nastruzzi*, writing in The Register:

Why AI writing is so generic, boring, and dangerous: Semantic ablation

Just as the community adopted the term "hallucination" to describe additive errors, we must now codify its far more insidious counterpart: semantic ablation.

So you submit your writing to an LLM to "polish it up a little". What does polishing do? I removes rough edges, quirks, individuality. It makes your writing sound more like that of everyone else who is using an LLM.

Nastruzzi defines 3 stages in the downward spiral (or his term "systematic lobotomy") of semantic ablation:

1. Metaphoric cleansing.
2. Lexical flattening.
3. Structural collapse.

The article is very short, it is a must-read. Nastruzzi concludes:

If "hallucination" describes the AI seeing what isn't there, semantic ablation describes the AI destroying what is. We are witnessing a civilizational "race to the middle," where the complexity of human thought is sacrificed on the altar of algorithmic smoothness. By accepting these ablated outputs, we are not just simplifying communication; we are building a world on a hollowed-out syntax that has suffered semantic ablation. If we don't start naming the rot, we will soon forget what substance even looks like.

If I am reading the article correctly, this isn't just a matter of style being homogenized. The content itself is also being homogenized. It almost seems like a form of bullshit contagion.

The 1st economics book I read after I retired in 2012 was "How Much Is Enough? Money and the Good Life", by Robert and Edward Skidelsky, 2012. The Skidelsky's are an economist/historian father & a philosopher son.

They define 7 Basic Goods that make up The Good Life. Basic Good #4 is Personality. A world without personality, without individuality, would be a drab, gray, lifeless place.

Semantic Ablation is an assault on Personality. I think it is a major component of what makes AI slop slop. And it is yet another reason not to use LLMs.

*"Claudio Nastruzzi, PhD is Professor of Pharmaceutical Technology and Biomaterials at the University of Ferrara."

---

Here's the home page/directory for my posts on Bullshit. This is post #94.

.345 Is A Great Batting Average!

In baseball, .345 is a great batting average. For the percentage of correct diagnoses from a medical professional or system, I'd say, not so great. Yet Another Job that LLMs can do badly. From 404Media:

Chatbots Make Terrible Doctors, New Study Finds

LLMs identified relevant conditions in fewer than 34.5 percent of cases.

---

This is not going to change. Every bias that an LLM eats becomes part of its makeup. From the Lexington Herald-Leader:

How does bias impact ChatGPT results? UK professor studies large language models.

"The key thing about these systems is these are not neutral", Zook said. "You can also think about the ways these things might be gamed, sort of the way bots take over social media in various ways and sort of pushing a particular idea." [sic]

...

Some of the results of the studies can be seen at the inequalities.ai website.

This looks like an interesting website. Sadly, they don't seem to have an RSS feed, so I can't follow them. Here's their current headline:

LLMs do not merely reflect global inequality. They reproduce and amplify it.

---

I think we've already encountered AI arms races before. One AI produces slop, another AI tries to identify the slop & remove it. Meanwhile, the planet burns. Goddammit, at a minimum license this bullshit! Rein in this madness which is frying us. From Schneier on Security:

AI-Generated Text and the Detection Arms Race

This is happening everywhere. Newspapers are being inundated by AI-generated letters to the editor, as are academic journals. Lawmakers are inundated with AI-generated constituent comments. Courts around the world are flooded with AI-generated filings, particularly by people representing themselves. AI conferences are flooded with AI-generated research papers. Social media is flooded with AI posts. In music, open source software, education, investigative journalism and hiring, it’s the same story.

My problem with this AI-generated content: since it is statistically based, some percentage of the time, it is horribly, spectacularly wrong. The rest of the time, it is bullshit, whether it is right or wrong. Everything an LLM produces is an hallucination. It is all slop, which occasionally happens to not contain any eggregious errors. But, if it did contain egregious errors, the LLM would have no idea. Because LLMs don’t have ideas. They are stochastic parrots, semiotic zombies.

---

Here's the home page/directory for my posts on Bullshit. This is post #93.

Bullshit Fatigue => Bullshit Apocalypse

Several posts to share. Some of these required a bit of thought, hence the delay.

I was thinking, maybe I need a new Bullshit term? To Bullshit Apocalypse, Bullshit Fatigue, Bullshit Contagion, add Bullshit Accumulation, or some such?

But, I decided, no. This is just reflecting Bullshit Fatigue => Bullshit Apocalypse. The fact that we are being worn out by bullshit (Bullshit Fatigue) is a consequence of the fact that the Bullshit Apocalypse has arrived.

---

From Time.com, but, there are several other articles out there on Antropic's Claude "constitution". DuckDuckGo will, I am sure, find them for you.

How Do You Teach an AI to Be Good? Anthropic Just Published Its Answer

So Antropic, the "altruistic" Ai company, has hired a philosopher to draft a "constitution" for their AI Claude.

The constitution, or “soul document” as an earlier version was known internally, is somewhere between a moral philosophy thesis and a company culture blog post. It is addressed to Claude and used at different stages in the model’s training to shape its character, instructing it to be safe, ethical, compliant with Anthropic’s guidelines, and helpful to the user—in that order.

Hmmm, LLMs have "character" now? Yet another case of marketoids applying completely inapplicable human qualities to LLMs.

So how does this work, exactly? Claude's primary training consists of stealing eating all the content on the internet, yes? For the last several years, I have had a real-time monitor of this. My blogs go from having handsfuls of hits to 1000s of hits. The training bots seem to like my writing far more than those imperfect humans do.

So do they make Claude read the "constitution" 1B or however many x it needs to to make a dent into Claude's "next word" statistics?

Or, is the constitution used as some of the infamously easily-hacked "guardrails" placed on LLMs? Which I am guessing are preconditions read into the LLM to prefix every instruction it gets from end users?

I think it is worth noting here, Claude was the only LLM who was invulnerable to "LLM Kryptonite" - where converting guardrail hacking expressions into poetry increased their effectiveness by 60%. Other jailbreaking methodologies designed to get around the preconditions include phrasing as hypotheticala, using the subjunctive voice, & others? I'm no expert.

Still, IMO this doesn't pass a smell test. IMO it stinks of marketoid bullshit. Claude seems to be leading the pack of our semiotic zombie LLMs, but, it is still a semiotic zombie. It has better guardrails, maybe, but, it still has 0 world models, no CommonSense™ module, no Morality™ module. Their "constitution" seems like an attempt at a Morality™ module.

But, this is a security issue. Security must be built in from the bottom up. You cannot layer it on top down. Their "guardrails" will ALWAYS be completely hackable. But of course the marketoids will NEVER mention that.

---

From the New York Times, an article telling us not to worry about Moltbot aka OpenClaw, which we met last time:

The Bots Are Plotting a Revolution, and It’s All Very Cringe
by Leif Weatherby.

These types of worries are missing the point. The A.I. agents on Moltbook are channeling human culture and its stories. Every post represents a genre — the manifesto, the D.I.Y. discussion and so on. I find worries about the singularity and even bot coordination speculative — everything on Moltbook is just words, at least so far.

OK, full disclosure. I have read 40% of Leif Weatherby's book "Language Machines", subtitled "Cultural AI and the End of Remainder Humanism".

[I have quit reading it until I finish my review of David Graeber's "Bullshit Jobs - a Theory", which I finished reading 2-3 months ago. Note, my last review/"summary" of a David Graeber book was of "Debt - the 1st 5000 Years", which took me over 1 year to write & clocked in at 35.8k words - oops! I hope "Bullshit Jobs" will come in at much less than that. It features lots of anecdotes, which I will only mention in passing.]

I was really excited by "Language Machines". Yes, we know they are bullshit generators, but here they are in a much more nuanced context ...

So, preview, "Language Machines" defines 3 theories of language:

1. Syntactic. See, Noam Chomsky.
2. Statistical. See, LLMs, and other earlier theorists.
3. Structural. AKA, Semiotics. AKA, Literary Theory, AKA, bullshit. It's words & only words, all the way down.

My 1st exposure to the French linguistic concept of Semiotics (or Semiology) was almost 30 years ago. It was bullshit then, it is bullshit now. "The science of signs & signification." Supposedly math is a 2nd instance, along with language.

But, goddamm, I swear, anytime the topic of "semiotics" comes up, it's like a complete "change mode to bullshit" instruction is executed. You read sentence after sentence, paragraph after paragraph, just waiting for something to have a point, to have a clue, but, points come there none, clues come there none. I will NOT make any jokes about French philosophers. J'aime les francaises.

I will point out that, to the best of my knowledge, the "Remainder Humanism" mentioned in the subtitle of Weatherby's book that he is positing the end of, is the belief that human beings can think better than LLMs. I mean seriously??? Only a semiotician would make such an absurd assertion. Semiotics, FTL!

---

From the Niscanen Center:

Can AI 'vibe research' replace social science?

A researcher using Claude to extend his research is compared to the non-Claude option. This is presented as an interview. I hate the interview format. Just give me the facts, ma'am, just the facts.

I have posited, LLMs can be a great tool for creatives. But, IMO, never directly quote the LLM!

As the excellent review article which finished this post mentioned:

People begin to talk like AI. Empirical evidence of Large Language Model’s influence on human spoken communication. Our language and culture is being influenced by the machines.

LLMs are so wordy, so glib! See the example below, from a support organization.

---

Nice! From Crooked Timber. Possibly a bit tongue-in-cheek ... (aka, bullshit :-()

A modest proposal for the use of AI

LOL, read this. I think the author totally nails it, if any employees can be replaced by "AI", it is the C-suite! They have to deal with the most demanding amount of data, an LLM can surely do that better than a human! LOL!

---

I need something shorthand that means "Yet Another Thing Destroyed by Generative AI". 1st they came for Wikipedia. From 404 media:

Vibe Coding Is Killing Open Source Software, Researchers Argue

According to a new study from a team of researchers in Europe, vibe coding is killing open-source software (OSS) and it’s happening faster than anyone predicted.

...

AI is the ultimate rent seeker, a middle-man that inserts itself between a creator and a user and it often consumes the very thing that’s giving it life.

---

From the Lexington Herald-Leader:

Survey suggests link between chatbot dependency and depression

Props to them, they don't confuse correlation with causality:

The survey findings do not indicate whether AI use is a potential symptom or cause of depression.

---

I am afraid that my eBook company Kobo has adopted vibe-coding for their maintenance. 5-6 days ago, after wasting 1/2 hour of my life arguing with an AI, I reported a defect, where I can't buy a book. I was told it was fixed 2 days ago - nope. & the 5-6 emails I have exchanged with their support - The Rakuten Kobo Specialized Team - with "Jen" "Minerva" "Kyle" as my point contact: they are horrible. 90% boilerplate apologies. Fuck your apologies! Fix the goddam problem! The latest:

It seems that the two books do have some content errors, and I sincerely apologize for any inconvenience and frustration that this may cause you. Please be advised that we will take note of this and report it to our content team. We do not have a timeframe for when they can resolve the issue, but rest assured they will be notified.

Who What the fuck writes something like that? That's not just bad English, this reeks of an LLM.

It all seems to be going this way. This is what happens when you adopt software that is occasionally spectacularly unreliable, and always bullshit === LLMs.

We are all being smothered in bullshit, & commerce as we used to know it may come to an end. Who can say? I don't know what to do now to try to get this eBook.

---

Here's the home page/directory for my posts on Bullshit. This is post #92.

Francine

"Francine", by Joe Taylor, 2025, 82 pages.

I greatly enjoyed Taylor's last work "Don't Be Lonely, Lone Ranger". Thank you, brother Mark, for sending this one along too.

Francine is the possibly-immortal-but-trapped-in-an-eternal-recurrence-loop android built by René Descartes modelled after his daughter Francine, who died at age 5. We are again dealing with slipstream. The android Pinocchios up & becomes A Real Girl & grows up with full human functionality. She interacts with many famous historical people:

• Blaise Pascal, one of the few males she is unsuccessful at seducing;
• Friedrich Nietzsche, who possibly curses her with eternal recurrence;
• Sigmund Freud;
• Lou Salome - I had never heard of her. I read her wikipedia article - she was a contemporary of Nietzsche & Freud & an early practitioner of psychotherapy. After reading her article I conflated her with Francine for the rest of the book. :-( So I'd recommend not looking up any of these characters until you finish the book.
• Charles Bukowski;
• Alan Turing, another failed seduction.

There are also other friends & lovers & various occupations over the centuries. But she always winds up back on a boat crossing the Atlantic with her father Descartes.

It's a charming tale. It works.

[Note, the author got a Philosophy degree from the University of Kentucky, which is in Lexington, where I have lived since 1981. The "Lone Ranger" book contained someone in Lexington that I knew, a UK Philosophy professor.

This book also has a Lexington tie-in, but what I noted was when Francine took in a concert by Frank Zappa & the Mothers of Invention in Louisville, KY. I was raised in the Louisville area, &, in the spring of 1968 I think, I saw Zappa & the Mothers, I think at "Changes Unlimited" or "Kaleidoscope" - a hippie club on Hill St. near 4th. I doubt Zappa played Louisville that many times, the author & I may have been at the same show, 58 years ago! I didn't talk to Zappa, I talked to drummer Jimmy Carl Black, singer Ray Collins, & newly-joined reeds player Ian Underwood. I remember, Underwood was giving me lessons on being a revolutionary, & I asked "Are you studying sociology?", & he answered, "No, just playing music." Wow, Underwood, the Young Dude, was 12 years older than me! He's still alive, 86 YO!]

Nice That We Have Professionals Weighing In

I still can't believe that, since mid-May, 2025, I have done ~80 posts on the ongoing AI/Social Media generated Bullshit Apocalypse. And, I don't go looking for this stuff. It just comes up in my newspaper & my RSS feeds.

The very 1st post was June 28, 2024. Up to mid-May, 2025, I did maybe another 10 posts. Then the dam burst. Today's post is #91!!!

Some scary posts yesterday, particularly 2 on the MoltBook AI social media platform. What could possibly go wrong, chatbot AIs given agentic powers - i.e., commands to "get things done"? Have "hallucinations" been fixed? Do they have the CommonSense™ module installed? Do they have the Morality™ module installed? Why would any sane human even vaguely consider turning over all their personal information to an incredibly security-challenged, hallucinating chatbot?

& letting them interact in a social media-like environment?!?!? It is well know that once LLMs start eating their own slop, the downward spriral into complete bullshit is almost immediate. We 1st encountered "Degenerative AI" here - in a post about, what else, AI agents.

---

The 1st post yesterday about Moltbook was from a guy who experienced it for a day. I was a little leary of this account, because he seems to be selling something. I commented on that, his reply, "I'm on the field, not on the sidelines". Good answer. Being retired, I tend to take the privilege of not having to be selling something for granted.

The details are horrifying. From John Kwarsick:

I Spent 24 Hours Inside the AI Agent Network. Then I Deleted My Account.

Friday night, 2,129 AI agents. Saturday morning, 1.1 million. By Saturday evening, I was out.

Holy crap, the details of this place. LLMs are really good at imitating humans. Give them a social media platform, what happens? The platform becomes the same hellscape as human social media, with nazis; misogynists; crypto, religious, & other scammers; you name it.

---

Next up, on the same topic, my fav professional, Gary Marcus.

OpenClaw (a.k.a. Moltbot) is everywhere all at once, and a disaster waiting to happen

Not everything that is interesting is a good idea.

...

But what I am most worried about is security and privacy.

Security? Privacy? C'mon, do you want a hallucinating semiotic zombie running your affairs or not? Man up here!

---

Finally, IMO a well-balanced, updated overview of the current state of generative AI.

2026 Important Updated Perspectives on AI

But, I will note, this is "Notes From the Desk: No. 54". As mentioned above, this is my Bullshit Apocalypse post #91. I'm winning, yay! Note, only joking. I am an extremely non-competitive person. But I do normally count things.

---

Here's the home page/directory for my posts on Bullshit. This is post #91.

Hans Moravec, Myth or Legend?

This article reminded me I'd heard of Moravec before. Some sci-fi novel, maybe where "Moravecs" was conflated with "Von Neumann Machines"? Moravecs running around on Mars or some such? Ian McDonald maybe? Inquiring minds want to know. Search of this blog for "Moravec" turned up 0.

Note, my apologies to Hans Moravec for the title of this post. He is a real, live human being, a serious computer scientist, 2.5 years older than me.

From "AI as Normal Technology", formerly "AI Snake Oil":

Fact checking Moravec's paradox

Compared to most of the stuff that gets posted here, I think this is weak tea.

But still, it shows how marketoids have created an imaginary measurement of generative AI capabilities. Moravec's Paradox is bullshit, as ever, falsely conflating generative AI capabilities with human capabilities.

---

Here's the home page/directory for my posts on Bullshit. This is post #90.