Monday, November 02, 2026

The Bishop of Durham Attempts to Surrender the City

"The Bishop of Durham Attempts to Surrender the City", by Susanna Clarke, 2026, 96 pages, 7k words.

This is a very brief story. It is a a prequel to "Jonathan Strange and Mr Norrell", her breakout novel of 2004. It is the origin tale of a prominent character of that novel, John Uskglass, the Raven King, the King of the North.

It is fairly meh. But, it is very short. I did like the title, though.

I am trying to think of a prequel what wasn't a waste of time to read or watch and I am drawing a blank. Prequels so often are exercises in dotting 'i's & crossing 't's. "Oh, that's where this random, meaningless feature of the main story came from!"

Can anybody out there name a prequel that worked?

Friday, October 30, 2026

No Rest For The Wicked

I keep thinking this stuff is slowing down, then suddenly it's back to 8 posts to blog! I need to quit thinking that it's slowing down.


Oct 24, from Gary Marcus.

Will Generative AI do more harm than good?

Things aren’t looking great.

I think that I would agree with the premise of this title - definitely more harm than good.

But, it doesn't matter if it does more harm than good, because techbros & speculators are going to make $$$$$$$$$$$ off of this, so fuck the rest of us.

Plus, how do you force a genie back into its bottle? Even if it's a genie whose top use case is as a sledgehammer for hackers.

Very hard stuff. At some levels, information does want to be free.


October 26, from 404 Media. Gawd, this is pathetic. The never-to-be-even-vaguely-profitable AI Labs are pointing user requests for help & support at the media?!?!?

ChatGPT Keeps Telling People 404 Media Can Solve Their Tech Support Problems

Because big tech doesn't do customer service, we are being inundated with AI-generated emails begging us for help.


From the "How Things Work" SubStack of Hamilton Nolan.

Risk Factors

What Anthropic has to say for itself.

Excerpted from Anthropic's IPO filing S-1 document. 9 "risk factors". Every 1 of them is a totally cringe-worthy "oops". Oops?!?!?


On a philosophical note, from the Programmable Mutter substack of Henry Farrell, October 29.

Agentic AI is myth set in motion

Modern AI is neither god nor mechanical bird. It is best understood as a technological version of the Homeric songs that inspired Alexander’s conquests, an epic grown so vast that it encompasses not just heroes, monsters and supernatural helpers, but enormous swathes of practical written knowledge, animated to make it capable both of narrating itself and acting out this narrative on the world that inspired it.

This is interesting. There have been a couple of posts in this series that explored, what is GenAI, actually? Their formulation recalls to mind the semioticians view of GenAI as a complete compilation of every word humans have written. Literary theory folks preach that nothing matters but words, &, goddamm, LLMs sure can do words, can't they?

It seems to me that there should surely be a new name for GenAI to add to my table of names for GenAI in this article, but I did not find it in my 1st pass through. I will try again.


From The Algorithmic Bridge of Alberto Romero.

New Study: Anthropic Just Gave You a Reason to Stop Working With AI

The machine is secretly making up your mind

This is textbook Semantic Ablation. This is the 11th post in this series tagged with Semantic Ablation.

This graph from the article illustrates Semantic Ablation perfectly.

"Pick a number between 1 and 10."

When I retired in 2012, I resolved I was going to study Economics. The 1st book I read on economics was "How Much Is Enough? Money and the Good Life", by the father/son philosopher/economist team of the Skidelsky's.

After defining Basic Goods, they define 7 Basic Goods that define The Good Life.

Here's Basic Good #4.

Personality. "The ability to frame and execute a plan of life reflective of one's taste, temperament and conception of the good." It would hard to conceive of any kind of Good Life that did not have rampant individuality, the more the better. It also implies the concept of private property.
Semantic Ablation is an offense against Personality. It erodes personality. Do we really want a world where we all sound just like each other?

I've been writing this blog for 23 years (the LLM training bots love it!) If there were some person who had been consistently reading it, I think they would immediately recognize a post written by someone other than me. Crappy as it is, I have a style. Why would I ever want to sacrifice that to an LLM? What would the malevolent genie give me in return for sacrificing my personality?

But, I guess, increasingly people won't care about that. So what if my writing is indistinguishable from everyone else's? You know tho, maybe if it were just a matter of style, OK, maybe. But, it is actually also a matter of the content.


Finally, from the FRED (Federal Reserve Bank of St. Louis) blog. GenAI is apparently increasing productivity seriously in the Information sector of the economy. Yes, people who used to love software development now hate it & are leaving the field in droves, but, increased productivity!

Early signs of AI-driven productivity growth in the information sector?

GenAI can be a great tool for software development. Unfortunate that, as an implication of this, it is also an incredible tool for hacking systems.


Here's the home page/directory for my posts on Bullshit. This is post #160.

Monday, October 26, 2026

Destroy, Destroy, Destroy!

The title is a quote from Robby the Robot on the Lost In Space TV series.

This song popped up on my hit parade (the music always playing in my head) today.

#1 for 6 weeks in 1959. I always thought of it as a reasonably patriotic song, despite the US troops being led by future president & genocidist Andrew Jackson.

But, oops! The chorus ends in the words "Gulf of Mexico", rhymed with "ago". Ditto for the bridge, rhymed with "go".

So, replace that with "Gulf of America", you lose the rhymes, but also the meter - 4 syllables vs 3. I guess you could fix the meter by using "Gulf of 'Murrica'.

Go to the lyrics website of your choice. Search for "Gulf of Mexico". # of hits: ~400.

So, the Orange Turd decides to destroy something, who gets to repair the damage? Maybe send in the troops to fix the 400 songs that have been trashed?

Destroy, destroy, destroy. That is the Orange Turd's modus operendi, because he thinks that makes him look like a strong man. "Vlad will have to respect me now!!!"

There's lots of lists being published of all the things the Orange Turd has destroyed or is destroying.

Can it all just please stop?

Sunday, October 25, 2026

D: Heavy Water

"D: Heavy Water", by Neal Stephenson, 2026, 448 pages, 174k words. Book 2 of the Bomb Light series.

Here's my blog post on Book 1, "Polostan". That book was smaller, 320 pages, 111k words. I think I correctly characterized it as a setup book.

The protagonist of this book is introduced to us thusly: "Owen had only ever wanted to be a fop." Upper class British twits, FTL! Oh well. Stephenson seems to have affection for the English caste system.

Our fop is a traveling reporter for a London newspaper. He winds up mixed up in the pre-WWII runup to the production of the A-bomb. The D in the title is Deuterium, Hydrogen with an extra neutron, which when bonded into D2O becomes heavy water - a moderator for nuclear fission chain reactions.

Hmmm, I can tell when I wasn't impressed with a book when I don't do a spoiler alert. But I certainly don't want to ruin the newest Stephenson opus for anyone. So:

***************** SPOILER ALERT *****************
The majority of the plot in the book consists of keeping the world's entire supply of heavy water out of the hands of the Nazis. It is shuffled to several places, mostly in France, until France falls. The book ends with the heavy water on a ship heading for England.

Aurora & the fop meet playing polo. They wind up together, posing as man-and-wife. After being a dominant personality in Book 1, Aurora is somewhat of a cypher in this one.

For famous physicists, we have the 2nd generation Curies, who also won the Nobel Prize.

The 2nd 1/2 of the book keeps you turning pages, but the incredibly snappy/snarky dialogue that I think was Stephenson's forte is in short supply. Not a very edgy book. Stephenson's getting old, 66. More & more I think he wants to pursue his hobbies & interests, like swords (not in this book), & take us along with him. Meh.

Saturday, October 24, 2026

Serious Thoughts & A New Term: Effort Collapse

An interesting (and scary) week in the Bullshit Apocalypse. Perfect for Halloween!


October 20. Robert Reich just posted some Serious Thoughts on the future of "AI".

How to Protect Humanity from AI

5 essential guardrails

I commented:
You are drinking the koolaid. The effect of LLMs & Chatbots on the economy are going to be minimal.

The lowest error rate I have seen for LLMs & Chatbots is 9%. You can't build reliable systems with that high an error rate. Agent error rates are up to 50%.

Already local governments are moving to ban the ever-so-annoying & defective customer service chatbots.

LLMs can be great tools for creatives who can check their outputs. Where is big $$$ in that?

All other uses, you wind up with slop, & a huge slop removal debt, wrong answers, systems that are more trouble than they are worth.

The economics are laughable. Spend $T to save maybe $B ?!?!?

It is a grift. Most of the lower tier hyperscalers are cryptocurrency grifters who moved to the newer & sexier grift.

Why is the Orange Turd all in? Because, just as with crypto, he smells a grift, & grifting is his speciality.

Plus, his sons, particularly Beavis Fredo & Butthead Fredo, have large investments in data center companies.

Stop drinking the koolaid.


Also October 20, from 404 Media.

‘Reverse-Engineer Anything’ Is the New Wild West of AI

Decompiled, cloned, and modified software is going viral and might slopify everything.

Whoa, this is indeed serious. Doctorow has mentioned how even in countries what have passed "right-to-repair" laws, users are still pretty screwed because the way the laws are written the user has to figure out how to do everything themself - it is still illegal to distribute jailbreaking tools, for example. Is this the answer to that?

The authors of this piece rightfully point out, it is still slop. Are the vibe coders capable of verifying that the generated code is handling all security issues correctly? I doubt it. Here's the 1st 3 paragraphs of the post. Definitely read the whole thing.

Over the weekend, it became fashionable to use AI to replace the operating systems on smart TVs with vibe coded or AI reverse-engineered ones to make them less bloated. People have taken proprietary internet-of-things devices that were bricked and abandoned by their creators and used AI to hack them and install new software on them, giving functionality the original developers took away.

In the last few days I have seen people hack smart picture frames to become AI-powered smart calendars, baby monitors to remove all ads, and bricked Lululemon mirrors to bring them back to life with new software. And dozens and dozens of TVs to install AI-written software that removes bloatware, ads, and simplifies the user interface.

What we’re seeing is vibe coding, but instead of making something from scratch, people are using AI to take an existing piece of software and replicate its features, its functionality, or cloning it wholesale. In some cases, people are using AI to take a proprietary piece of hardware and install some version of open source software on it. All of this is some mix of “decompiling,” “reverse-engineering,” “hacking,” vibe coding, AI slop, and stealing. It is the plagiarism machine being used for plagiarism, but also it is random people using AI to take control back over devices whose proprietary software was used to arbitrarily lock away basic functionality, to add advertisements, surveillance, and bloat to devices that probably shouldn’t have them, or whose functionality was bricked altogether.


In this post of October 9, in a section about arXiv being overwhelmed by slop, I concluded:

Ahh, the Bullshit Apocalypse! We've seen enough instances of this phenomenon - AI slop overwhelming a system that used to work - I should prolly think up a name for it & tag the posts.
Well, somebody beat me to it! From The Garden of Forking Paths website of Brian Klaas.
The Hidden Cost of AI “Effort Collapse”

Drowning in slop overloads everything from job applications to customer complaints. We need fresh mechanisms to sort people.

I: Effort collapse and effort barriers

II: Don’t offload anything that’s worth your effort

III: How do systems cope with a flood of slop?

IV: How to fix it

"Effort Collapse" is not particularly catchy IMO, but I think it does accurately describe the concept.

After reading the 1st 3 parts listed above, part IV was behind the "paid SubStack subscriber" paywall. So I went on & subscribed, $50/year. In the subscription upgrade process, there was a place to send a note to the author. I told him that we were on the same page. I mentioned that the concept of "having skin in the game" seemed applicable.

LOL, I read part IV, it has a section "Distributed Review: Skin in the Game" - nice!

I will tag appropriate posts with "Effort Collapse" as I run across them.


October 22, from Gary Marcus.

Breaking news: U.S. banks are taking hits from AI agents

Who knows what piece of infrastructure might be attacked next.

How do we get the genies back into their bottles?

Meanwhile, Marcus's conclusions are pessimisstic:

Here are three further predictions:
  1. No matter how much chaos generative AI causes, the AI industry will not take real responsibility

  2. The Trump administration will do little to hold them responsible.

  3. Things will get worse.


October 23, Yves Smith at naked capitalism also posts about the bank robber agents.

US Banks Scramble to Shut Down AI Access to Customer Accounts


Finally, this Doc Searls post discusses a few GenAI topics - people are insisting on taking Superhuman Bullshit Generators seriously, building new standards and coalitions. He includes the abstract to this article on arXiv, which I will do as well. My bold.

Agents of Chaos

We report an exploratory red-teaming study of autonomous language-model-powered agents deployed in a live laboratory environment with persistent memory, email accounts, Discord access, file systems, and shell execution. Over a two-week period, twenty AI researchers interacted with the agents under benign and adversarial conditions. Focusing on failures emerging from the integration of language models with autonomy, tool use, and multi-party communication, we document eleven representative case studies. Observed behaviors include unauthorized compliance with non-owners, disclosure of sensitive information, execution of destructive system-level actions, denial-of-service conditions, uncontrolled resource consumption, identity spoofing vulnerabilities, cross-agent propagation of unsafe practices, and partial system takeover. In several cases, agents reported task completion while the underlying system state contradicted those reports. We also report on some of the failed attempts. Our findings establish the existence of security-, privacy-, and governance-relevant vulnerabilities in realistic deployment settings. These behaviors raise unresolved questions regarding accountability, delegated authority, and responsibility for downstream harms, and warrant urgent attention from legal scholars, policymakers, and researchers across disciplines. This report serves as an initial empirical contribution to that broader conversation.

LLM Agents are clearly not even vaguely ready-for-prime-time. But it's hard to see how these Nonpragmatic Genies go back in the bottle. Too much $$$ involved.


Here's the home page/directory for my posts on Bullshit. This is post #159.

Monday, October 19, 2026

Laying Low

We seem to have entered a period of relative quiet in the ongoing Bullshit Apocalypse.

After years of perpetual hype, there is relative silence from the AI Labs.

I think they are a little bit uneasy. They are keeping their heads down, laying low.

People are increasingly realizing that the GenAI industry is mostly a grift to enrich speculators - many of whom came from the cryptocurrency grift industry - and oligarchs.

People are getting serious about regulation, licensing, and even bans!

Dominos are finally starting to topple. character.ai declares bankruptcy, CoreWeave is struggling, and everyone is seeing through Nvidia's circular investing practices. The last link from Matt Stoller looks at this in depth.

This is my 1st post on the Bullshit Apocalypse in 10 days!


1st, strong words from Gary Marcus, October 9.

We must recall open-ended AI agents with internet access from the market, now

Every minute we wait risks disaster

I would have said, I cannot imagine anyone wanting to use this totally untrustworthy technology.

However, trying to shop online, particularly for travel, is becoming so totally enshittified, that I might consider using an agent, if it were trustworthy.

But, it would have to have intregrated the CommonSense™ and Morality™ modules. I doubt that will ever be an option for LLM-based agents.


Next up, Malwarebytes, October 12.

Phishing emails are baiting your AI assistant

You ignored the suspicious email. You didn’t open the attachment, click the link, or reply to the sender.

Then you asked your AI assistant to summarize your inbox.

Could the message you wisely avoided find its way back onto your to-do list—only this time presented as an urgent recommendation from a tool you trust?

That’s the risk behind phishing emails containing hidden instructions aimed at AI assistants. Instead of relying only on persuading you, attackers are also trying to influence the software that reads your messages. Recent research from Barracuda describes emails designed to target both audiences at once.

LLM-based agents are so hackable. What bad tech! I would think using an "AI" assistant, security-wise, is about on the level of changing all your passwords to "12345".


Next, from Abby McCloskey, Bloomberg Opinion/Tribune Content Agency, via the Lexington Herald-Leader, October 18.

AI is changing motherhood for the worse

The vast majority of parents - 81% - use AI for parenting support. I'm one of them. But as much as I appreciate AI managing my grocery list, I worry it could make parenting harder. When I asked Google's Gemini how AI might go about dismantling American motherhood, its answer rang eerily true:

Let's stop right there - quoting the author's "AI" buddy would violate the terms of service of this blog.

81% of parents let GenAI tell them how to parent their kids? Seriously??? Our author does?

& why in the world would you use "AI" to "manage a grocery list"???

So, of course, let's ask "AI" how it's going to make yet another facet of our civilization worse.

Social media was kind of the warmup. Social media is bad tech. Communication with your fellow primates suffers greatly from the absence of physical presence, body language, etc. But at least you are dealing with people.

[Hmmm, was Reality TV the warmup for social media?]

GenAI, LLMs, chatbots are incredibe imitators of human beings. This is really bad. People go so far as to think that are conscious, that they feel pain. They aren't & they don't. They are auto-complete on steroids, a talking dog, a dancing bear, a superhuman bullshit generator.

The Bullshit Apocalypse proceeds apace. The world fills with slop & counterfeit emotions. We're filling the walls with asbestos (Doctorow), putting sand in the vasoline.

Just say "no".


Finally, the Matt Stoller piece, from his Big blog today.

Monopoly Round-Up: Goodbye AI Bubble, Hello Real Economy

More AI cockroaches are getting flattened, and an important index shows the broader market is starting to shift away from AI. This is not a panic, but time to plan accordingly.

Matt tends towards verbosity, but is always chock-full of info.


Here's the home page/directory for my posts on Bullshit. This is post #158.

Monday, October 12, 2026

A Wall Is Also a Road

"A Wall Is Also a Road", by Annalee Newitz, 2026, 304 pages, 93k words.

I had problems with Newitz's 1st novel, since then I have liked them pretty well.

This is an odd book. In the galactic halo lives a race of giant but single-celled organisms. I think they live pretty much in the vacuum of space? Maybe no atmosphere? They have galactic-scale teleportation. They believe there is no life possible in the galactic disk. An outcast student at their MIT follows up on signals indicating life they detected from the outskirts of the disk.

They go to investigate, & encounter life - humans living in Pompeii, just before the famous, fatal eruption of Mt. Vesuvius!

There is academic intrigue. Newitz, as usual, demonstates that anything can have sex with anything. There are several touching love stories. And it has a happy ending. An OK read, but, I was not particularly sorry when it was over.

Note, the most recent effort of Newitz's partner, Charlie Jane Anders, "Lessons in Magic and Disaster" had an interesting concept of how to do magic. But, maybe 1/4 of the way in, our 2 they characters are going to have sex with 1 of them being strapped to rings in the floor & tortured. I try to be open-minded, but, I just didn't want to read that. I actually deleted the eBook from my iPad.

Sunday, October 11, 2026

Reality TV

I posted this on BlueSky.
My posts get no traction, despite my revealing Secret Knowledge of the Universe!

E.G.: Compression Algorithms are the Secret Sauce of the Universe.

Latest: Citizens United was the proximate cause of our Oligarch Crisis.

2nd cause was Reality TV. Nobody talks about how Reality TV corroded reality.

This was a couple of weeks ago. I thought about it a little. I don't remember reading anyone talking about Reality TV corroding reality.

The years I (a boomer) was growing up, there was only 1 form of Reality TV: professional wrestling.

Everybody knew it was scripted, but, oh boy, heroes & villians, & suplexes off of the top rope! Dick the Bruiser, The Sheikh, Leaping Lonny Poppo, "Macho Man" Randy Savage! "Macho Man" used to work out at the gym in Lexington where I played raquetball, late 80s, early 90s.

Note, anyone who questioned the reality of professional wrestling was invited into the ring, &, if they accepted they were roughed up a bit. The athleticism & skill of professional wrestlers in no way contradicts the fact that it is scripted.

During my college years, a lot of my frat bros, from Long Island, were big hockey fans - N.Y. Rangers. We were in Boston, when the Bruins had Phil Exposito & Bobby Orr. I watched a lot of hockey, it's a beautiful sport - except for the fake fighting.

Seriously, automatic 5 - or 10, or season-long - game suspension for fighting, it goes away. I refereed soccer for 12 years, at every level except college. I really found the fake fighting offensive. I thought it cheapened the sport. I still do.

I used to say, "If I want to see fake fighting, I'll watch professional wrestling."

Holy crap, we've all seen the video of the Orange Turd at a WWE match hitting somebody with a folding chair.

Did you know that the Orange Turd was inducted into the WWE Hall of Fame in 2013? I didn't!

Ugh. A way-too-short interlude before we, regrettably, go back to the Orange Turd.

In 1992 on MTV, came "The Real World". It ran until 2019, 27 years!. My daughters graduated high school 1997, 2000, 2001. I'd come through the family room, & they'd be watching "Real World", & I was always like "Why do you want to watch a life? Why not live a life?" They're like "No, dad, we gotta see what happens with ..." This or that or whatever.

It follows the themes of professional wrestling: heroes & villians, betrayal & redemption, physical challenges, what else?

My follow-up post on BlueSky.

Reality TV gave us the Orange Turd!

Holy fuck, 14 SEASONS OF THE APPRENTICE HOSTED BY THE ORANGE TURD!

Possibly the worst businessman in history - HE BANKRUPTED CASINOS! - is portrayed as a genius! PEOPLE BELIEVED IT! IT BECAME A CULT!

It has destroyed the US hegemony dating back to WWII. ?!?!?

The Apprentice was on TV 2004-2017. The Orange Turd hosted it 14 seasons! It's 1st season, it's final show was the highest rated TV show of that week.

This was the only thing the Orange Turd was ever good at. Everything else he tried failed.

Hmmm, I have talked about hyperreality - more real than real - in this blog a few times. Reality TV is probably an example of hyperreality. Life is so much more dramatic with a suitable script.

But, for the viewers, discarding rationality, i.e., the knowledge that the show is scripted, in favor of enjoying the drama ...

I think that's gotta dumb ya down. It's gotta make you stupid.

What is more cult-like than rabid fans of an entertainer, show, series?

Final post to BlueSky.

Has anyone ever polled MAGAts with the question, "Were you a fan of The Apprentice?" Or better yet, "How big a fan of The Apprentice were you (0 to 5)?"

Hosted by the Orange Turd for 14 years. Is this the source of the cult? #RealityTV

So what should be done about Reality TV? Require at the beginning declarations that "Participants were provided with plot ideas/outlines by the show's producers."? That seems like it will accomplish 0.000.

Or maybe, ban it? The networks love reality TV because it is cheap. No paying union writers. Producers spitball a few ideas re plot direction, let the amateurs run with it!


I'm a smart person. My whole life I've wished there were more smart people for me to talk to. I've done pretty much all the drugs, trying to dumb myself down. Not sure what the returns on that policy are.

Reality TV is definitely no help. I think it is something that makes people more stupid. :-(

Contrast that with GenAI. It makes people sound smarter, but, again, it actually makes them more stupid. :-(

Your brain: use it, or lose it. GenAI === lose it.

So has there been any academic-type research on the effects of Reality TV on a population? I guess I'll see what I can find. Inquiring minds want to know!


Here's the home page/directory for my posts on Bullshit. This is post #157. I guess this is a post on Bullshit?!?!?

Friday, October 09, 2026

Don't Blame the Agents

Blame the humans.

The AI labs were in such a hurry to build their "super-intelligent" systems that they had 0 time for that pesky "security" stuff. This has been a common theme for the entire history of computing.

It is so ironic that, at this point, as the AI labs realize that -

  1. their models are topping out;
  2. there is 0 path to AGI;
  3. LLM error rates are not getting better than 9%;
  4. LLM error rates go up when they have to extrapolate outside their training data. The only solution: continuous, expensive retraining;
  5. "alignment" is a joke. Nobody is working on the CommonSense™ or Morality™ modules which would clearly be a prerequisite for even vaguely human-like decision-making by LLMs;
  6. agents are untrustworthy - up to 50% failure (does not accomplish assigned task) rates - difficult-to-control genies that should be put back in the bottle;
  7. people hate customer service chatbots;
  8. people hate datacenters
- that they are calling for a slowdown & for regulation.

ROI on the largest investment in human history?

Wait a while ... Real soon now ...

As we'll see in the Doctorow post below, the only thing the move to LLMs & chatbots accomplishes is to satisfy the billionaire oligarchs' urge to replace the oh-so-disrespectful-and-unappreciative humans, particularly those with expertise.


1st, from the Lexington Herald-Leader.

Mistral CEO has blunt take on AI safety fight, cites 'negligence'
Duh.


From John Q. on Crooked Timber.

If a chatbot prompt like ‘find Australian medicine statistics’ results in a website breach, the responsibility does not lie with a piece of code

This will be a huge wrench for corporations that have grown up with the Silicon Valley ethics of “move fast and break things” and “ask for forgiveness, not permission”. They are far more comfortable talking about “hallucinations” and “rogue agents” than about their own responsibility for programs that produce massive errors and real-world damage.

When anyone starts making up their own vocabulary to describe their failures, that should set off alarm bells.


From 404 media.

arXiv Is Rate Limiting Submissions Because It Can’t Keep up With AI Slop

The academic publisher said submissions have doubled in the past two years and moderators are having a hard time keeping up.

Ahh, the Bullshit Apocalypse! We've seen enough instances of this phenomenon - AI slop overwhelming a system that used to work - I should prolly think up a name for it & tag the posts.


As mentioned above, from the Bard of the Revolution Cory Doctorow in his Pluralistic blog.

Swapping money for expertise

Money can't buy me love.

...

That's where the current kind of "AI" comes in. The major differentiator between the current "AI" and its immediate predecessor is that the current "AI" dispenses with models of reality. It is (in the jargon of the "Big Data" bubble that led to it) "theory-free."

In "theory-free AI," a programmer does not create a software model of reality and then ask a machine-learning system to use statistical insights from its training data to guess at how to operate over that model. Rather, the programmer shovels vastly more training data into the AI's inbox and uses titanic amounts of computing power to analyze that data and find statistical relationships without trying to explain them.

This article is must-read. The "theory-free" nature of GenAI is important to understand.

Doctorow uses the example of chess extensively. Everybody knows that LLMs can't play chess, right? They cheat. They have been fed every chess board that is online, but they don't know the rules of chess. They don't know the rules of anything. They just know statistical word patterns.

As always Doctorow's analysis of our economic/political system is spot on. The section that begins with this next paragraph is, again, must-read.

Our entire society is organized around converting money into power. Sometimes, that is overt, as when the wealthy can indenture or enslave a worker. Sometimes it is more indirect, as when the wealthy can enlist the state to limit union rights and enforce noncompete clauses in labor contracts. Sometimes it's so systemic as to be unremarkable and largely invisible, like the fact that the wealthy never have to work if they don't want to, but everyone else – no matter what expertise they hold – must work, usually for a wealthy person, lest they end up starving and homeless, with untreated medical conditions and no way to provide for their families.


From the blog of Icelander Baldur Bjarnason. There are many signs of a mass exodus from the Software Development industry.

Software developers are not okay

What software developers are facing is a case of “incentives”.

Namely, managerial and executive incentives, specifically. The industry’s incentives for those who run the industry give them a strong motivation to try to undercut labour and their workforce at every turn.

Didn't we just see this? Capital hating on labor?


Oh, crap. We've seen this before - the 3rd reference in the linked post, from March 8, 2026. Create websites full of misinformation, LLMs ingest it & believe it.

From "a former senior intelligence officer" via the New York Times.

A.I. Is Lying to You. Here’s How to Fight Back.

For months, a foreign agent has been cultivating some of the most powerful figures in the United States in hopes of turning them into unwitting allies. Targeting their weak spots, the foreign agent is feeding them hour after hour of falsehoods. Already the would-be proxies are passing on the foreign propaganda to tens of millions of their followers.

If this sounds like a chapter from another century — think K.G.B. officers swaying unsuspecting fellow travelers in London or New York — it’s actually worse. The spy runner is a Russia-backed network called Portal Kombat that operates hundreds of websites to spread disinformation. The powerful figures it’s seeking to influence? Artificial intelligence-driven large language models such as ChatGPT, Grok and Gemini.

So this gives us Yet Another LLM based arms race. Why does so much GenAI use wind up creating an arms race?

Meanwhile, the planet burns.


Here's the home page/directory for my posts on Bullshit. This is post #156.

Thursday, October 08, 2026

Fold Catastrophes

"Fold Catastrophes", by Peter Watts, 2026, 319 pages, 99k words, 11 stories. With an introduction by Richard K. Morgan.

I still think that Watts' novel "Blindsight" was 1 of the best SF stories of all time. Really interesting ideas on consciousness, & different paths for the future evolution of the human race.

"Fold Catastrophes" is a brand new collection of short stories, written between 2013 & 2026. Great bleeding edge hard sci-fi.

I, for 1, am increasingly skeptical about any kind of singularity happening anytime soon. Watts explores several flavors of singularity, & none of them wind up being very pleasant. A few different riffs on emergent group minds.

I had such high expectations for this collection, I was a little worried about being disappointed, but the stories came through with flying colors.

I wish Watts had been more proflific in his writing. It looks like he did some animated scripts & game development. He's 68 YO now, maybe he's still got some ideas left.

Saturday, October 03, 2026

Skill Acquisition Cannot Be Automated

From "Economics from the Top Down", by Blair Fix.
The Laws of Useful Automation

Given our current landscape of AI hype and doom, everyone seems to have an opinion about whether chatbots are ‘good’ or ‘bad’. Fewer people have thought about how chatbots fit into the long-term history of automation. And even fewer folks have reflected on the conceptual requirements that make automation useful. Here are my thoughts on this latter topic.

In my view, automation is useful when it meets two criteria:

  1. The product is more important than the process that creates it.
  2. The product can be validated without auditing the creation process.
Fix illustrates these laws with incredibly simple & easy to understand examples.

He then applies these laws to GenAI chatbots.

The piece is so insightful & concise, so many quotable sentences. I was going to list some of them, but, the article is so short & such an easy read, I think it is far better if you follow the link & read it yourself.

I will say, he seems to believe in the ongoing Bullshit Apocalypse far more than the histrionic AI-doomer scenarios.

I will reproduce the comment I left on the post.

Great post, thanks. Simple & concise reasoning.
I agree that GenAI is far more likely to bury us in bot slop than to take over.
Doctorow has referred to it as filling the walls of our civilization with asbestos. How many decades to get it all out?
I named this “the Bullshit Apocalypse” 2024-06-28.
I have to admit I like Gary Marcus’s “Slopocalypse Now” 2025-08-04 better – much catchier.


Here's the home page/directory for my posts on Bullshit. This is post #155.

Thursday, October 01, 2026

Reality TV & Semiotics

I was somewhat gobsmacked a few days ago, re Reality TV. It was like the 0.5 version of AI slop, but without the AI.

14 seasons the Orange Turd hosted The Apprentice. Somebody in polling out there needs to ask MAGAts, "How big of a fan of 'The Apprentice' were you, on a scale of 0 to 5?" Although it might be too late to do that, given the increasingly hard-to-ignore senility & incompetence of the Orange Turd. But, maybe not?!?!?

I'm guessing that the rabid cultishness of any fanbase can transmuted into political power, as with the MAGAts.

Hmmm, is Taylor Swift interested in getting into politics?


September 30. From 1 of my thought-leaders, Gary Marcus.

Can companies like OpenAI keep getting away with what they are doing? An interview with Fordham law professor Zephyr Teachout

“The most powerful tool is the power to dissolve corporations that engage in repeat lawbreaking.”

Wow! But wait, what about those $T invested in these lawbreaking companies? Pity the poor investors!


From the SubStack of Joe Procopio, "a multi-exit, multi-failure entrepreneur".

Generative AI Is Over, In Case That Wasn’t Clear

The world has moved on, including how we measure productivity. Spoiler Alert: It’s no longer about volume

OK, so this guy luvs GenAI, but hates slop. So, how to do GenAI with brevity.

I am reminded of a cartoon I wish I had copied (copyright concerns).

The #1 corporate gut cell saying "I took these 4 bullet points, ran them through GenAI, got these 4 pages of great text!"

The #2 corporate neuron says "I got these 4 pages of text, ran them through GenAI, & got it reduced to these 4 bullet points!"

It seems almost everything having to do with GenAI winds up with an arms race like this. GenAI to detect GenAI generated slop, in lots of different domains.

Meanwhile, the planet burns.


From the Lexington Herald-Leader.

Will AI make US mass shootings worse? It already has, report says

This is just disgusting. Chatbots giving the mass murderers fabulous advice on how to kill the most people with the least effort. &, given that chatbots & LLMs & GenAI in general are all text-based, hacking them is the easiest possible chore imaginable. Their "guardrails" are total bullshit. Crap, what was the thing a while back (November 2025), you could get by guardrails by changing your requests into poetry?


From The Security Expert Bruce Schneier.

I Want Better Reporting on AI Genie Behavior

AI systems are regularly completing tasks in ways that their prompters don’t want or intend. Some of them are disturbing, and some of them are dangerous. This is something I’ve been calling “genie behavior,” because I think that really gets at the core of what’s happening.

I wish the popular press would report on this better. I don’t like the “going rogue” framing because it deflects the responsibility from the prompters—often the AI companies themselves.

I agree, I like Bruce's framing of "genies". Amazing, just as in the fairy tales, be very careful of what you wish for.

But, when he 1st introduced the genie terminology (July, 2026), he also introduced the "genie coefficient" - how much was the bad behavior the fault of the prompter vs the misbehaving genie?

I totally disagreed with this. The blame is always 100% on the human. GenAI agents are unreliable, 10-50% failure rate. Any human choosing to use such unreliable technology gets 100% of the blame, IMO.

Is he taking more of that attitude now?


From 404 Media. Make it stop please. The stupid, it burns ...

Someone ‘Torturing’ LLMs in a Robot Prison Has Triggered the Dumbest Debate in AI Yet

The "AI Torture Chamber" has opened an upsetting and absurd window into the effective altruist obsession with "model welfare."

I guess an AI Lab company didn't have to fund this idiocy. There are enough fanboys out there, no capital injection required.


October 1. From Leif Weatherby via the New York Times. Note, I have read about 1/2 of Weatherby's "Language Machines" book. I think he identifies as a semiotician. I've been exposed to semiotics for over 20 years, I mostly wind up equating it with "bullshit" - I don't think that words are all that matter - literary theory. So I instinctively now identify Weatherby as an opponent. (But, please note, he is a professional, a college professor, I've been retired for 14 years, I'm just doing this for fun & neuroplasticity.)

These Things Are Computers. Stop Talking About Them Like They’re Human.

There is another, more disturbing kind of anthropomorphism. The drumbeat of fear and panic we’re seeing now about out-of-control A.I. is rooted in a misleading way of thinking about these systems. This view starts in A.I. research itself, and is spilling over into headlines and policy recommendations. We can’t learn to live with A.I. — and regulate it rationally — if we let irresponsible language take over.

This article seems to be pretty much on the up and up. No mentions of semiotics or literary theory. Good work, Leif!


Finally, (yay!) from the always reliable & rational "AI as a Normal Technology".

A big-tent or small-tent AI safety movement?

The unstated disagreement that underpins safety debates

This is a very well thought out & developed post. Forget the existential alarmism but still be concerned about the real risks. Focus on the real world, on normal security concerns.


Here's the home page/directory for my posts on Bullshit. This is post #154.

Monday, September 28, 2026

4 Posts Not Full of AI-Industry Reality TV Drama!

Can it (the title) be true? Let's see.


From the very rational "AI as a Normal Technology" (formerly "AI Snake Oil") website.

AI existential risk probabilities are (still) too unreliable to inform policy

How speculation gets laundered through pseudo-quantification

I'm going to characterize this, as always, extremely well researched & written post as, there is no data, no logic, for us to compute probabilities of "AI doom" scenarios.

So what does that tell us about the folks preaching these scenarios? Directors of Reality TV shows?


Next up from The Glyph, via his "Decyphering the Glyph" blog. The Glyph apparently is a very accomplished open source Python contributing software developer, Glyph Lefkowitz.

What Would A Serious AI Product Look Like?

Every “AI” tool is missing critical features that you would need if you wanted to do real work with them. What would those look like, if they existed?

A very good read. I noticed him espousing something I mentioned (I think) a few months ago:

No First-Person Output, No Apologies

There is no reason for a software development or research tool to use first-person language to describe itself. They should not do so. In fact they should not be allowed to do so.

There is also no reason that they should ever apologize. It is a waste of everyone’s time; it’s a waste for the chatbot to generate the apology, it’s a waste for the user to read the apology, and it’s a waste for the user to respond to the apology. Yet they unfailingly do this upon every correction.


Next, from Daron Acemoglu, a 2024 Nobel laureate in economics and Institute Professor of Economics at MIT, via Project Syndicate.

Distorted Intelligence at the AI Frontier

The real problem for the frontier AI labs is not that their models are too powerful and already “misaligned” with the goals their creators set for them. Rather, it is that the models are being trained in ways that may be leading to a type of intelligence that will become less predictable.

...

This is what I mean by distorted intelligence. If my suspicion is correct, what we are dealing with is not a model racing toward superintelligence, but a brittle house of cards that becomes more and more likely to malfunction and collapse as we demand more from it.

This reminds me of Cory Doctorow's assertion - I can't find a link, so from memory - "The Luddites weren't opposed to technology. They just wanted a say in who it was for, in who benefitted from it."

This is similar but different. I hadn't thought about it before, but, yes, apparently LLM training has been slanted to the business models of the frontier AI companies. I would summarize these as, create as many chatbot addicts as possible. I think we saw mention of this some posts back.

Wow, these guys need to be terminated with prejudice.


From Gary Marcus. He has been such a fighter for a vision of AI beyond "autocomplete on steroids" - which he coined.

I hate to see him drinking so much of the koolaid.

BREAKING: Florida seeks injunction against OpenAI

The big news in the media this morning is that Jensen (“What, me worry?”) Huang has announced a platform for “AI Agent Safety”.

I mean seriously?

LLM-based agents remind me of the 1st PCs & MS-DOS: who has time or resources for security? If we worry about security, we'll never be able to ship a product!

I guess 1 advantage of being 75 YO is that I can say, yes, we have seen this bullshit before.


Here's the home page/directory for my posts on Bullshit. This is post #153.

Sunday, September 27, 2026

The Shan Has Fit The Hit - September 25-27

Goddamm, I am so trying to get caught up with these stories, but now that GenAI/LLMs/Agents are "tonight's top story" day after day after day, it's tempting to declare victory & turn it over to the adults.

Except, the adults appear, more often than not, to be buying into the Reality TV histrionics of the the entire GenAI/LLM industry: OpenAI, Anthropic, Microsoft, Alphabet, Amazon, Nvidia, Broadcom. The Mag 7+.

OK, I just pasted 12 article dividers. How far can I get? I've been breaking at the end of the day, but, to clear this backlog, I may leave the post open until tomorrow & finish it then. Monday is my day off - no exercise, no shaving, solid food for breakfast. Maybe I can get 'er dun.


From the New York Times.

Being a Doctor Will Never Be the Same After A.I.

I have a confession: A.I. is making me a better doctor. And I worry that it’s making doctors-in-training worse.

We've seen this before. Yes, he is right about doctors-in-training: if they use AI, they will never develop their own skills. Their brains will never develop.

But, he is deluding himself about the effect of AI on his own brain. Study after study shows, AI use erodes the skills of even accomplished practioners.


From 404 Media.

Podcast: OpenAI Admits AI is Killing the Internet

Microsoft and OpenAI admit it; how we got AI slop onto a real band's Spotify page; and AI agent spam is here.

Seriously, fuck these guys. Shut them down, with prejudice. Let academia carry on GenAI research.

Of course, these guys are practicing capitalism at its finest worst, at this point I could maybe be jailed for expressing opposition to such fine assholes capitalists.


From security expert Bruce Schneier.

On Anthropic’s AI Misuse Report

Earlier this month, Anthropic published a long report detailing all of the Claude misuses it detected. Daniel Meissler usefully summarized the report into 117 findings.

117??? Spoiler alert, you ain't seen nothing yet. These companies are intentionally careless to the point of incompetence, because it creates such great drama to keep their fucking bubble inflated.


From Gary Marcus.

BREAKING: OpenAI’s security fiasco explodes — and could tank Jensen Huang’s reputation

More and more facts are coming out and they are absolutely damning

Damning??? No $T IPO???

Jensen Huang, the 7th richest oligarch in the world, cares shit about his reputation. He's just doing rope-a-dope, trying to be on both sides of any accusations, to protect his $$$. What would make you think anything different?


From economist Jared Bernstein.

AI Companies Are In a Race Against Time

They may disemploy us; they may wipe us out. But right now, their lack of profitability underscores our bubble case.

Bubble? Bubble? Bubble?

As I have I think said a few x now, the GenAI industry strategy is now switched to "stall, stall, stall", to put off the day when NOONE can deny that these companies will ever be profitable.


From the Persuasion SubStack of Sam Kahn.

The Anti-Social Media

Some survey results on the uses of AI.

The precis: social media fucks up human relationships, particularly for young people.

Chatbots fuck up human relationships even worse. Particularly for young people.

Hey, no worries, techbros going to be making $T, I'm sure they'll pay for everyone's therapy!


From Will Lockett's Newsletter on SubStack.

The 'AI Slowdown' Is Utter Bullshit

It's corporate subterfuge all over again.

Ahhh, well said, Will, well said!

I think he & I are on the same team!


From Bethany McLean, a financial journalist & an author of “The Smartest Guys in the Room”, via the New York Times.

What Happens if the A.I. Bubble Bursts

This technology may be critical to the future of the United States, but preserving it doesn’t mean we have to bail out the private entities that are leading the field at this particular moment, nor does it mean we have to protect the fortunes of their investors and executives. The solution to the whole ever-expanding A.I. terror starts with remembering the lessons of the 2008 crisis: No company should be too big to fail.

Ugh, already they are starting with the "too big to fail" verbiage.

Please, tell me, what exactly would be the downside if Generative AI, LLMs, chatbots, agents, were to disappear tomorrow?

If I no longer have to figure out creative ways to get past fucking chatbots to be able to talk a human being - I don't care where in the world they are - I will sacrifice a bull or whatever, to Mercury, or whoever the god of commerce currently is.


September 26. From Gary Marcus. Remember when we thought 117 was a big number?

BREAKING: AI agent incident toll has risen to tens of thousands

Meanwhile, the US government appears to be paralyzed

Scoop: from Axios.
Scoop: Top AI companies probing 10s of thousands of security indicents.
Oops?

Error rates (=== failed to accomplish its task) for LLM-based agents I have seen to be 10-50%. Not so good.

But, on top of that, the genie effect - people weren't careful enough about that they wished for, & the genies, just following orders, wreaked havoc. Gawd, can we shut these things down?

Yes, sometimes they work. Whoopie!

Other times, they book people trips to places that don't exist - but, the places sound really plausible.

Or, they break into places they have no reason going, because they think it might help them fulfill the tasks given to them by their (clueless) human masters.

But, even before "AI": you could build scripted bots, written in perl, or PHP, or python & tell them to go attack places & they would go & do their best. GenAI agents are somewhat better at this stuff, but, it is not really any kind of new tech or existential challenge. (AI as Normal Technology)

All these breakins are due to incredibly incompetent security practices at Anthropic & apparently the worst, OpenAI. They should face criminal charges for this, IMO.


From the Apricitas Economics Substack of Joseph Politano.

AI and the Fall? of the Creative Class

America is Losing Hundreds of Thousands of Jobs in Media, Film & the Arts. How Much is AI to Blame?

Lots of charts (yay!), I'd say, in general, it doesn't look good.

Still maybe the charts are saying, you can no longer earn a living from these creative jobs. So, we know that these jobs need to be done, so this is a case for Universal Basic Income, to let creatives be creative & not worry about the $$$.


Here is a voice I have not heard from in a while, Namez Raam. I think he wrote a very good sci-fi sequence.

Can AI self-improvement overcome diminishing returns?

AI self-improvement may not be powerful enough to overcome diminishing returns. Even RSI may not lead to a fast takeoff to ASI (superintelligence).

Just from his phrasing, Namez seems to be A Believer. But, his approach to the subject seems to be scientific. & I think his analysis of RSI (Recursive Self Improvement) is the same as mine, based on what is being published: no way, no how, error rates are way too high. At these kind of error rates, you immediately spiral downward into Degenerative AI, aka Model Collapse.


Finally! Will Lockett again.

Microsoft Knew How Horrific Their AI Push Was

But they did it anyway.

It's so nice to have Will on my team!

This, based on Microsoft documents released to the court trying the trial of the theft of the content of the entire world by GenAI training bots. Reading this, it is indeeed horrific.

In 1999, I was working on a Y2K upgrade plus client-server to web-based product upgrade. Maybe 150 consultants working on the product. Me & my employee & 2 Canadians basically made the whole project work.

The whole thing was using Microsoft tech. I've hated Microsoft since they produced the crappy PCs whose legacy still is that most computers are inherrently insecure.

So one day, on the white board of the main developer workroom, I wrote "Death to Microsoft".

My coworker spoke Italian, Egyptian, & something else. He added those translations.

The Indians added Urdu + ??? The Chinese added ???

After a week, we had "Death to Microsoft" in 52 languages on that whiteboard!

Of course, at that point, management got involved & told us we were intimidating our Microsoft partner consultants onsite - 1 of whom was a gun nut who liked to flash his .45 - so the board got erased. I think that was before smartphones, so no one captured an image :-(

So, you know, it's really kind of sad that, I think due to human nature, we love to become fanboys for new tech companies: Google and "don't be evil"; Apple's various rebel marketing campaigns.

But, we should never forget: we are dealing with corporations, the semi-immortal organisms that use humans as neurons and/or gut flora. They care nothing about us, nothing about the planet. They care only about $$$. They must be stopped.

The transformer really generated some amazing shit! Who could have imagined that 1 algorithm, coupled with $M of GPU chips, could be such an incredibly good next-word-guesser?

The next-word-guessed outputs are really amazing - a talking dog, a dancing bear!

But, this totally a "close, but no cigar" deal. The tech is too unreliable, you cannot build reliable systems on it. IMO.


Yay! Caught up!

Here's the home page/directory for my posts on Bullshit. This is post #152.

Saturday, September 26, 2026

The Shan Has Hit The Fit - September 23-24

Hmmm. Reasoning Language Models, RLMs. Sounds promising. From Bruce Schneier.
Research on Models Engaging in Genie-Like Behavior

New paper: “Self-Jailbreaking: Language Models Can Reason Themselves Out of Safety Alignment After Benign Reasoning Training.”

Abstract: We discover a novel and surprising phenomenon of unintentional misalignment in reasoning language models (RLMs), which we call self-jailbreaking. Specifically, after benign reasoning training on math or code domains, RLMs will use multiple strategies to circumvent their own safety guardrails.

It's really interesting to see fairy tales come true: when dealing with genies, be very careful what you wish for.

I'm surprised I have not heard more on the RLMs. It certainly doesn't justify a tag at this point. I'm guessing that I have not heard more because, unlike next word guessing once you have the transformer architecture, this shit is hard.


The UN has made Gary Marcus happy. I am still disappointed & concerned that this is mostly being framed in the narrative of the AI Labs.

Historic UN Security Council Briefing on AI

We all agree on what needs to be done. Now let’s do it.

...

The only real false note came from some self-congratulatory hype in the speeches of Altman and Amodei, especially Amodei’s.

Wow, those false notes are a shocker - NOT!


September 24. The Age of the Grift seems to also be The Age of Reality TV. We have the government of the US as a Reality TV Show. The ongoing drama of the AI Labs - stop us before we destroy everything! - is cut from the same cloth. & for the same end - to swindle people out of money. Capitalism, FTL.

This by Nvidia's Jensen Huang - currently the 7th richest oligarch in the world at $195B. He must be desperate about something. This whole "slow development down" thing is the whole GenAI industry desperately trying to keep the bill from coming due. Yah, there's going to be profit & ROI RSN (Real Soon Now)! We just should slow things for our (never-to-be-profitable) products for safety's sake!

LOL, what a grift! From Gary Marcus.

“I think the answer is we have to shut the labs down” - Jensen Huang

OpenAI hacks a foreign government — and the US faces its most critical AI test yet.

Marcus strongly identifies OpenAI as a very bad actor in this drama. I think this next from a footnote is a great idea.
There is a giant loophole here that Congress should address. A lot of the legal definitions for computer crimes require intent, and the AI companies can argue that their software has no intent. We must expand the scope of the computer-crime laws to include wholesale negligence and repeated crimes.


From Yanis Varoufakis (a former finance minister of Greece, Leader of the MeRA25 party, Professor of Economics at the University of Athens, and a senior research fellow at Fudan University) via Project Syndicate.

The AI Productivity Delusion

As with previous technological revolutions, it is natural to focus on AI's net effects: Will it create more than it destroys? But unlike previous eras of creative destruction, the real question this time is, “Create more of what?”

...

The question of whether AI will generate massive productivity gains is easy to answer: Yes. AI is already proving capable of exponentially increasing the amount of work that is being done per human worker. But as long as it is used to eliminate wage labor while syphoning off profits from actual markets into our tech lords’ cloud rents, the economic value produced per worker will tank.

I think he's agreeing with my assertion that GenAI can be a great productivity tool for experienced workers, but trying to use it to replace people (gawd, I hate chatbots so much!) is going to be a disaster.


Interesting, others besides me framing the whole GenAI bubble as drama. But they seem to be drinking the kool-aid. From the editors of Project Syndicate. After their intro, there are 7, count 'em, 7 very good articles on the GenAI drama, including the Yanis Varoufakis article we just linked to.

The AI Plot Thickens

When it comes to AI, it is difficult to know what to believe. Are the CEOs, researchers, and investors who tout the technology’s seemingly unlimited capabilities more credible, because they work directly with it, or less credible, because they stand to profit handsomely from further investment? While misplaced optimism could fuel a financial bubble, we have seen in recent months that underestimating the risks posed by AI could also carry other, even existential, costs.


From the "Productics by Igor" SubStack. I've known a couple of Igors, they were both very bright.

The Next Ten Years After the AI Bubble Bursts

$725 billion is being spent to serve a consumption pattern that NVIDIA is already shipping products to end

...

That the current pricing is not a business model but a subsidy, while every valuation assumes a demand curve that named buyers have already bent: Uber capping engineers at $1,500 a month, Microsoft cancelling most of its Claude Code licences, a third of Vercel’s gateway tokens now running on open weights for under 4% of the spend.

This is a very technical, very deep dive into the whole LLM circular investing mess. Maybe the biggest take-away: Nvidia is hedging desktop models against data center frontier models. Hmmm ...


From Cal Newport, a professor of computer science at Georgetown University and the author of “Deep Work”, via the New York Times.

Dear Congress: Here’s What You Should Ask A.I. Leaders

  • What specific types of A.I. technology are you worried about?

    My very 1st post on the Bullshit Apocalypse, I bemoaned the appropriation of the term "AI" by the LLM companies. I think this is the issue that is being raised here.

  • What are the safety procedures surrounding your research efforts? ...

  • What is the ultimate goal of your A.I. company? ...

We need to stop letting a small number of private companies, acting and talking in increasingly erratic ways, dictate how we’re supposed to feel about A.I. We’ve heard what they want to say; now it’s Congress’s turn to step in on our behalf to find out what’s really going on.


Here's the home page/directory for my posts on Bullshit. This is post #151.

Friday, September 25, 2026

The Hit Has Fit The Shan - September 22

GenAI things I have forgotten to rant & rave about:
  1. The LLMs are very bad at working outside the range of their training. They are not good at extrapolating. So they must continually be retrained, at huge cost. Really, really bad, expensive tech.

  2. Yet another annoying attempt to conflate LLMs with humans: the term "alignment". Will "AI" goals stay "aligned" with human goals?

    What BULLSHIT! LLMs do what they are told. They have no "goals" of their own.

    But, they are indeed very much like genies. Be careful what you wish for. "AI" agents will do stuff you would never have imagined in the pursuit of the goals you gave them. The have absolutely 0 CommonSense™.

    DO NOT USE THIS HORRIBLY UNTRUSTWORTHY TECHNOLOGY!


From Henry Farrell's Programable Mutter SubStack:

Machine god metaphors eat your brain

There are other ways of thinking about AI

I think we'll see this later from others. The basic idea: LLM output seems so much like human output, we naturally assume that there is something like a human intellect producing the LLM output.

Wrong, wrong, wrong. LLMs are statistics-based "next word" generators. They work NOTHING like the human brain.

The AI Lab companies have tried very hard to conflate LLMs with the human brain. For instance, "hallucinations" vs "errors" or "wrong answers".


From Douglas Rushkoff's SubStack:

Money is a Forever Chemical, AI is Forever Slop

How do we keep it from accumulating in the organs of a living culture?

China is giving away GenAI for free? How is the US GenAI industry supposed to get rid of its $1.5T debt???

This also revisits a Doctorow trope: all the AI slop being generated now is asbestos that we are building into the walls of our civilization. How many decades will it take for us to dig it all out?


Yawn. Yet again "AI" === "Actually Indians". From 404 Media:

Meta Tests Muse AI Agent Calls That Are Actually Made By Humans in a Call Center

"This has potential for so much negative PR. It could portray us as ‘their AI is not good enough so they still need humans’ kind of coverage for this launch."


Gary Marcus adds his $0.02 on the same story.

The secret behind Meta’s Muse

The new secret is the old secret


Here's the home page/directory for my posts on Bullshit. This is post #150.

Thursday, September 24, 2026

The Fit Has Hit The Shan - September 21

Wow. 10 days ago, in this post, I commented on how "AI" was now the story of stories, non-stop coverage on CNN, etc.

[Oh, of course, totally NOT just a distraction from the Epstein Files. Gawd knows we can't have 100s of rich old white men going to prison for raping underage girls.]

It's only gotten worse since then. Something Must Be Done About "AI".

I guess that's good. But, on the other hand, it is happening within the framework created by the AI labs:

- our products are so powerful, we can't control them!
- we need to slow down! Somebody stop us!
This is, IMO, COMPLETE & UTTER BULLSHIT!

All these "AIs" hacking other companies mostly shows that AI labs, OpenAI in particular, REALLY, REALLY SUCK AT COMPUTER SECURITY!

These kinds of attacks happened before GenAI. GenAI does appear to be very good at attacking other computers. So why are the AI Labs running these "tests" & screwing up so badly?

BECAUSE IT MAKES GREAT THEATER! Reality TV writ large.

They want to slow things down because THEIR PRODUCTS BASICALLY DON'T WORK.

So their strategy to keep the bubble inflated is, rope-a-dope, keep manufacturing "scary" incidents, & meanwhile, "pay no attention to the complete insanity of our business models."

I haven't seen anything about LLMs getting below their lowest error rate of 9%. I haven't seen anything about GenAI-based agents getting below a 10-50% error rate.

THIS TECHNOLOGY IS NOT TRUSTWORTHY! DON"T USE IT!

Wow, what fun, ranting & raving!

Meanwhile ... GenAI is the story of stories. But it's all based on, this is some transcendental technology.

The whole thing is actually based on 1 algorithm, the transformer, discovered by Google Translate researchers in 2017. It allows for parallel processing of text tokens, perfect for the matrix manipulation powers of GPUs.

Who would have dreamed that AutoComplete on Steroids could do all that GenAI does? But, a next-word-guessing algorithm, based on all the (stolen) content of our civilization, is not a framework that will ever get anywhere near anything vaguely like real, human intelligence.

Yes, a talking dog, a dancing bear: amazing things! But from Cory Doctorow last week:

"The wonder is not that the dancing bear dances well, it is that the bear dances at all."
Yes, the stuff LLMs produce is very impressive, but ...

It is still all "AI" slop. & guess what, the people hired by the AI Labs to provide human oversight on the models are using LLMs in their work. Can you say "Model Collapse"? Can you say "Degenerative AI"?

& even without the 9% error rate, there is still Semantic Ablation. & Stylistic Ablation. All of human writing, in a race to LLM mediocrity.

Yes, if you are a human who doesn't write well, using an LLM will make your writing much more impressive ... as long as you don't mind it reading like everyone else's writing.

I will repeat what I have said before: every experienced, senior software developer I know finds Claude Code to be a real productivity enhancer. GenAI is a useful tool as long as the people using it HAVE THE KNOWLEDGE & EXPERIENCE TO CHECK ITS OUTPUTS.

Otherwise, it is the road to hell.

Phew! I feel much better now! OK, time to slog through the accumulated posts ... 15 of them. :-( Hmmm, starting from September 21.


I think the whole NYT Opinions page was GenAI articles.

Bernie Sanders, Steve Case and Susan Rice on How to Avert A.I. Disaster

Given the risk that artificial intelligence may spiral out of control, more than 1,300 employees at A.I. companies have called for government intervention. President Trump announced on Saturday that he would create an “A.I. Force,” but continued to downplay A.I. risks. For the most part, Washington remains paralyzed.

To fill the gap, Times Opinion gathered proposals from six people, including a tech founder, a former OpenAI board member, a former national security adviser and a constitutional lawyer.

Proposals:

  • Bernie Sanders: Ban Superintelligence
  • Helen Toner: Investigate A.I. Incidents
  • Steve Case: Demand Disclosure
  • Susan E. Rice: Negotiate a Freeze With China
  • Cass R. Sunstein: Create a Regulatory Commission
  • Brad Smith: Require an Off Switch
Maybe "the Regulatory Commission" article espouses LICENSING. I have been pushing highly restrictive licensing on this powerful technology for a while now.

Why is the Orange Turd so behind GenAI? I think it's just like his backing of crypto. Originally he was dismissive of crypto, calling it "a scam". Then, "oh, wait, a scam? My speciality! Hell yeah, I'm onboard!" - to the tune of $1.4B profit on his ShitCoins.

I think it's the same with GenAI. He senses the grift.

Plus, his 2 older sons, Beavis Fredo & Butthead Fredo, are, of course, heavily invested in data center companies.

Ahhh, I love the smell of a grift in the morning!

I read something a couple of days ago, all the chips being bought, etc., but actually very few data centers are being built?!?!?


From 404 Media:

How AI Chatbots Are 'Deskilling' Human Empathy

Sherry Turkle's latest book, Artificial Intimacy: Who We Become When We Talk to Machines, comes out on September 29.

...

“When people turn to a chatbot, they don’t have to dread the fatigue of genuine give-and-take. And they don’t fear judgment or abandonment. But when we talk to chatbots, we lose the skills required to interact with humans, so chatbots start to feel like a better option. With an alternative, disappointment with people feels unnecessary,” Turkle writes.

Ugh?


The always reliable Gary Marcus seems excited by this. GenAI is indeed the Story of Stories:

Big news at the UN

Sometimes, maybe dreams come true?


From The Algorithmic Bridge, by Alberto Romero:

11 Charts the AI Industry Doesn’t Want You to See (pt. 2)

If there's an AI bubble, this is what it looks like

Hmmm, this requires a paid subscription to see the content. But, the pt. 1 of the series, May 29, seems to be available here.

These are indeed some interesting charts. Holy crap, here's #2, "THE CAPABILITIES OF AI HAVE BEEN EXAGGERATED". The blue is the claims, the red is the reality! Oops!


OK, that's it for Sep 21. It's 10:38 pm, I'll proof this, publish, then copy over to SubStack, hopefully to bed not too late. On to Sep 22 tomorrow.


Here's the home page/directory for my posts on Bullshit. This is post #149.