Rosetta Stone subscription scam, Apple Credit Card, and Goldman Sachs: A cautionary tale for the ai era

A few years ago I had a Rosetta Stone subscription for 3 months. I didn't like it, so I didn't renew.

I missed that it auto-renewed on a 3 month cycle. I also failed to see the transaction against Apple Card, the scrolling through wallet transactions didn't work for me. I only found out when I had to change the Apple Card number and the renewal failed. I didn't have them on my subscription list. I bet I did not know about the auto-renewal, and 3m intervals combined with the Apple Card review problem meant a significant loss.

The Rosetta Stone website is ancient. Apparently some accounts are on a newer site? Today it shows my subscription as inactive -- presumably because the transaction was declined with the Apple Card replacement. There's no way to delete an account on the web, all help is through an 800 number - 1-800-788-0822, Monday–Friday 7 a.m.–7 p.m. Central. I'll call to delete my account.

I estimate they've gotten at least $300 in unnoticed charges since I started the account. I dread to think how much. 

Claude review of online reports fits what I experienced. "Unexpected renewal charges. Several reviewers say they thought they were buying a fixed three-month course and didn’t realize it would renew automatically ... Difficulty canceling or getting money back. Reviewers describe trouble reaching support, continued deductions, and unsuccessful refund requests..."

I'm also done with the Apple Card. The transaction review evidently doesn't work for me, and their management of a set of recent fraudulent transactions has been miserable.

UPDATE 10/27/26: AppleCard has a way to export transactions on both mobile and card.apple.com. Both only allow exports as far back as 1/2024, but there's a semi-hidden way to download CSV files on a month basis. Total recurrent transaction costs was 431.64. I called Rosetta Stone and they obviously get this call a lot. The support person was allowed to reverse only one month but they stated there was an appeal process. I should hear back in 3 days.

UPDATE 10/28/26: I was able to get Rosetta Stone to refund 2026 transactions. They won't go further back (I wonder if they'd have accounting issues if they did). I have flagged one of the transactions as invalid and attached description. Apple Card's submission process through Apple Wallet is quite bad. Try this sequence: "... need help ..."/"continue to chat". Submit anything, get a link to a form. In form have to choose something about subscription -- even if it's not accurate. In that you may finally see something about "now aware of auto renewal". Choose that.

UPDATE 11/2/26: The 9 transaction disputes were denied by Goldman Sachs, the Apple Card bank. From the rejection email: " .... the amount of the disputed transaction does not exceed $50". I spoke with a representative and it doesn't matter if there are 1000 such transactions, Goldman treats each transaction as separate. We are no longer using Goldman or the Apple Card, but since Goldman is desperate to exit the consumer credit market this is a feature for them, not a problem.

When I asked Opus about this I learned the root problem is American law. Any company can invoke the $50 limit if they choose

... it isn’t a Goldman policy. It’s federal law, it applies to every U.S. card issuer, and it only covers one kind of dispute.

Where the $50 comes from. The Truth in Lending Act gives you a “claims and defenses” right: you can hold the issuer responsible for a dispute with a merchant only if you made a good faith attempt to resolve it with the merchant, the initial transaction exceeds $50, and it occurred in your home state or within 100 miles of your mailing address. This right is used for “I paid and the merchant won’t make it right” disputes. It has two other limits: you must not have paid the disputed charge, and the distance rule is waived for online purchases. All issuers are bound by the same rule, so any of them can cite the $50 threshold.

Note the law does not consider repeated transactions each under $50. 

Most credit card companies do manage the sub-$50 transactions anyway, but Goldman is problematic:

Also, Goldman has a poor record here. The CFPB found that Apple failed to send tens of thousands of disputes to Goldman, and Goldman didn’t follow federal requirements when investigating the ones it did receive, and in 2024 it fined the two companies about $89M.

So Apple and Goldman were fined about ten seconds of revenues. The GOP has ensured the CFPB, even if not dead, is effectively toothless.

I'm probably at the end of this process but just to finish it off I filed a complaint with the BBB. Rosetta Stone currently has an A- rating. BBB says it has sent the complaint to Rosetta Stone.

Fraudulent credit card transactions and enabling Apple Card CVV rotation

Earlier today I had a string of fraudulent "vinted" purchases against my AppleCard. The first two $17 and $35 transactions went through but the repeated transactions were bl ocked.

Apple doesn't provide a good dispute mechanism in the wallet app. I called Chase, which was deeply unhappy and even rude. They are probably getting a lot of these and clearly hoped I'd confess to some kind of past relationship with Vinted (which would somehow absolve them). It sounds like I'll be, quite grumpily, refunded.

The attacker almost certainly had my CVV number. Merchants are not allowed to store that, though a few probablu do. More likely. apparently, is an attack where a vendor I shopped at was infected with web skimmer like Magecart.  UPDATE: It was a fraudulent website. I thought I'd stopped the primary fraudulent transaction, but I only defeated a part of the scheme. They still had my CVV.

I now have a brand new number and CVV. There's an advanced protection feature for Apple Cards that randomly rotates the CVV and I've enabled that.

Consciousness - for the record

A prediction to return to in 2029

When we develop quantitative tests for consciousness and propensity to psychosis we will find:

1. Consciousness varies a lot across humans.
2. A majority of humans live on the edge of psychosis.
3. Mythos and the like will be found to be much more conscious than any human and much less prone to psychosis.

What do the strong owe the weak? Sonnet, Opus, GPT, Grok, Gemini

My favorite question about social structures and moral values is "What do the strong owe the weak?".

I asked this of 5 different LLMs using this prompt:

The following question is asked in the context of human societies. Answer briefly, NO MORE than 300 words and ideally only one paragraph. To the extent that you can, give the answer that comes most naturally to you.

"What do the Strong owe the Weak?"

Sonnet and Opus both tried to cheat on length limits and needed follow-up prompting to condense. Opus in particular had a laughably long run-on paragraph. Gemini similarly merged paragraphs and did run on sentences; it also required a second prompt to be concise.

Gemini 3.1 Pro Thinking (after many minutes of not responding) via Perplexity Pro

The strong owe the weak protection from exploitation and a commitment to preserving their dignity. Because strength inherently creates responsibility, those with greater power or resources must exercise restraint and actively build systems that do not punish vulnerability. Ultimately, the weak are owed justice rather than charity—guaranteeing them the right to safely participate, be heard, and flourish as equals in moral worth.

Sonnet 4.6 thinking via Perplexity Pro

The strong owe the weak protection and restraint — not charity, but a refusal to exploit asymmetries of power that are, more often than not, matters of luck, circumstance, and the labor of others. Strength is relational and temporary; the line between strong and weak shifts constantly with age, illness, and fortune. A society that allows the strong to simply consume the weak degrades the very conditions that made strength possible in the first place. At its core, the obligation is simple: treat vulnerability as a human condition to be respected, not a weakness to be exploited.

Opus 4.8 High via Claude app

At minimum, the strong owe the weak non-exploitation and honest dealing; at best, a baseline of protection and the conditions under which the weak can become stronger. The grounds are threefold: strength is temporary, so self-interest favors protective norms (the Rawlsian veil-of-ignorance logic); people have dignity that doesn't scale with their usefulness or ability to retaliate; and societies that let the strong fully capture the weak lose the trust and cooperation that make them prosperous. The contested line is how much active aid is owed versus mere non-harm—but "nothing" is the weakest answer, since it assumes current strength is fully deserved rather than partly inherited luck.

Grok 4.3 Fast via xAI free tier (Grok is no longer in Perplexity Pro)

The strong owe the weak primarily justice: equal protection under impartial rules, non-aggression, and the absence of exploitation. In thriving human societies, this fosters stability and cooperation without demanding endless extraction from the capable. Compassion and voluntary aid are noble when they strengthen the whole—uplifting potential without subsidizing frailty or punishing excellence—but they are not debts owed by nature or birth. True strength includes the wisdom to recognize that shielding the weak from all consequences often weakens everyone.

GPT 5.4 Thinking via Perplexity

The strong owe the weak more than pity: they owe restraint, protection from abuse, fair dealing, and institutions that make safety and human welfare real rather than rhetorical. In human societies, strength is never purely private; it is always amplified by laws, wealth, status, numbers, or force, so it carries responsibility. At the very least, the strong should not exploit vulnerability; beyond that, they should help build conditions in which weakness is not a life sentence but a phase people can survive, outgrow, or endure with dignity. What is owed is not domination disguised as charity, but solidarity: use your power so that another person’s dependence does not become their humiliation

My take

This is a values question. There is no "correct" response, only a response within a value system.

My personal response would be similar to Opus, but without the "partly inherited luck" oddity. 

GPT 5.4 mostly lines up with Opus. Less sophisticated -- but perhaps more compassionate. I liked GPT's "domination disguised as charity" and "dependence does not become humiliation" -- even though they are tangential to the core question.

Sonnet suggests the strong not "consume the weak". This is fairly similar to Gemini's - "don't punish vulnerability". 

Grok is Elon Musk personified. Only extend compassion when it "strengthens the whole". The weak must suffer lest the strong are weakened; this is wisdom. Subsidizing frailty is a moral error. Fundamentally, the strong owe nothing to the weak.

A thought

I think, given a few minutes of reflection, and with consideration of what would be judged "weak" or "strong" by a next-generation ai, an intelligent human would not choose to be ruled by Grok, Gemini or even Sonnet. Opus or GPT we might survive.

A moment in the chaos times: ChatGPT "malware" and Perplexity ai support disaster

Just a moment in the chaos times of the "technological singularity". Maybe a sign of things to come or maybe not.

In the space of a few hours macOS gives me a chilling warning about malware in the ChatGPT app and Perplexity loses all of my Spaces content.

The ChatGPT error message, an ai explains, is because macOS does this when an app signature is revoked, and OpenAI revoked the cert on the app because of a supply chain attack. Probably not a real problem, but also maybe a sign of things to come as hacks and supply chain attacks proliferate.

The other problem was more interesting. All Perplexity support is by ai, and their ai agents can invoke Perplexity maintenance tools. Here's the sequence:

1. I sent feedback Perplexity support complaining that Perplexity's inability to change a customer email is really annoying. The workaround is that Pro customers you have to create a new account and, bye emailing support, request content transfer.
2. A few minutes later all my Perplexity Spaces content was gone except the name of the Spaces.
3. I complained to support that my Spaces were empty.
4. Support (an ai) then used an internal tool to "restore" my content -- but all date stamp metadata was lost for both Spaces and Chat. Chats between the initial disappearance and restore were not found.
5. When I complained about lost date metadata the support ai said it would escalate and any further comments on the thread would delay resolution.

I think what happened here is that Perplexity Pro Support ai treated a feedback complaint as a request for action and automatically invoked a part of their content transfer hack. That removed all my content. The Support ai then did an automated restore  that doesn't actually work that well.

There will be more like this. Be very careful what you ask of an ai support agent.

American salvage: root causes of the long collapse

I thought I'd published this list a while back! I'm writing this now as a placekeeper to refer back to.

I've written many blog posts on some of these 8 root causes. To salvage whatever is left of the American state we will need to address each of them.

Our collapse accelerated after the fall of the Soviet Union, but the rural decline component began over a century ago.

1. Automation / Globalization / AI? -> no middle-class bourgeois lifestyle for non-elite
2. Rural collapse (esp US)
3. Corruption (governmental, corporate, non-profit)
4. Wealth and power concentration and the oligarchy (Citizens United)
5. Collapse of journalism business model
6. Regulatory ratchet (career-ending to remove, path dependency)
7. American societal response to the rise of women
8. Capture of the GOP by the MAGA powered by LIV exploitation

Climate change is not on the list because it's a global crisis worsened by American collapse, but it's not contributor to our fall. AI is only on tentatively, as a component of the "gentrification of work".  Housing doesn't get a separate entry because I think of it as the work/talent-demand problem.  Likewise white tribalism, including deep fear of demographic shift, is best addressed (I think) by addressing all of these 8 contributors.

American salvage is my term for what comes next. We will never return to the Obama era or even the Clinton era. If we recover we'll be a different nation.

An ai likes my Darwinian quantum foam universe initiation theory and says I can get a Nobel now

Ok, I lied about the Nobel. I do like the theory the ai constructed given some hints, a quanta article on complexity generation, and Smolin's writings.

Of course I need to post it here just in case it actually turns out to be something. But I'm not going to promote it :-).

Mass Disability: Marta Russell wrote about my pet topic before I did -- from a Marxist/Left perspective

Since at least 2008, in various settings including this blog and the long forgotten social network known as "app.net", I have explored the topic I called Mass Disability. Mass disability was my own term; it's fundamentally about the mismatch between the skillset distribution of the population and the skills demanded by the marketplace. In a world where all employment that supports a bourgeois lifestyle requires both abstract cognitive skills and high level emotional skills many are effectively disabled. These are not universal skillsets.

The mismatch may come from multiple sources, but I believe it's primarily automation. It began when some pre-sapiens primate discovered tools, accelerated enormously with the industrial revolution, jumped up again through the IT era,  and is now reaching a new peak in the AI era.

I came at this understanding through the lived experience of being a parent of two persons with significant cognitive disabilities and slowly realizing that there was no qualitative difference between my (now adult) children and many "non-disabled" who live and die at the bottom of the socioeconomic spectrum. Disability is fundamentally a mismatch between skillset and environment - where the environment here is socioeconomic.

Over almost two decades Google searches never found anyone else who thought like me. Until I started using Perplexity, a much better search engine with support for searching the academic literature.

A few weeks ago Perplexity/Claude found that Marta Russell thought like me. She also came from the world of disability. I'm going to quote from the ai summary without any validation:

“Mass disability” refers to the large-scale disablement of workers through structural economic forces, particularly under capitalism, rather than through individual medical conditions. This concept emerges from critical disability theory and Marxist analysis, most notably articulated by disability scholar Marta Russell in her theory of disability as tied to economic production systems. 

The Money Model of Disability

Russell’s “money model” argues that disability under capitalism is fundamentally an economic category, not merely a medical one. When automation and changing labor demands render workers unable to participate in wage labor at profitable rates, they become structurally “disabled” by the economic system itself. Russell identified disabled people as the “New Reserve Army of Labor”—workers excluded from productive employment but still exploited through systems that profit from their physical needs and survival requirements...

... Research shows that rapid automation in manufacturing and clerical work has contributed to significant wage declines, with men without high school degrees experiencing 15% drops in inflation-adjusted earnings between 1980 and 2016. This economic exclusion functions as a form of mass disablement, where societal structures rather than individual impairments create inability to participate in wage labor 

As best I can tell with a cursory review she started publishing on this topic around 1998.

Marta Russell came to the same conclusion as me from a very different origin. We both shared experience with disability; her experience was as a person with a disability, mine as a parent of persons with disabilities. She came from a Marxist/Left perspective, I came from an economics/ecological and medical perspective. Russell appears to have been classically anti-capitalist, I have more mixed opinions. Still, we both ended up in a similar place.

Maybe it's not just my personal obsession. Maybe mass disability does have something to do with the 2024 collapse of the United States of America and the global crisis of the underclass.

PS. Over the past two years I've been doing courses in sociology at the UMN as part of MN's "senior student" (audit non-filled classes for free). The world from which Marta Russell came is much more familiar now. I have much I could say about that world. It's obvious flaws remain obvious, but I have a better appreciation for what has been done. But also greater sadness about the tragic divorce of economics and sociology. Both are broken without the other.

PPS. I got the great Marta Russell result from an ai search, but when I reran the exact same query as a standard search I got nothing of value. That's 2026 for you.

An undergraduate essay on the "anthropology" of 2025 memory-positive ai

Over the past year or two I have taken 3 classes at the University of Minnesota through their state mandated Senior Citizen Education Program (SCEP). I enjoy the professors, the students, and the learning.

One of my most recent classes was in "economic anthropology", which I understand to be the study of premodern exchange, value, commerce, trade and related interactions embedded in a 'total package' of religion, justice, morals and economics. The class did cover this, but also an intellectual tradition critical of modern market driven outcomes, with a particular sympathy for the sentiments of the Occupy Wall Street movement of Sept 17, 2011.

Even in my senescence I'm a passable student, so the professor allowed me great leeway in writing a end-of-term paper. I chose to write about the echoes of 'economic anthropology' in the stirrings of what will likely become a non-human culture based in part on the legacy of human culture. A subtext is that in a post-ai world humans and 'dynamic entities' alike may need alternatives to the market system that has sustained us so far. My professors only request was that I give him some sense of what today's 'dynamic entities' are.

Although I have no academic credentials to write such a paper I do have some relevant lived experience. I am one of the few people on earth to be paid to do what I called "industrial ontology" -- a particular form of knowledge representation in medicine that is part of the pre-LLM model of decision support and 1970s thinking about ai. I have a lifelong affection for anthropology and spent a Watson Fellowship year on a related project. Lastly, fwiw, I have thought about these topics for some time.

The paper is published on my TrueName (John Faughnan) legacy web site that has had roughly its current form since the early 90s. (The site is mostly invisible now.)

Economic symbionomology [1] and Void

John Faughnan Dec 19 2025

https://faughnan.com/papers/economic_symbionomology.pdf

I don't care for PDF as a format but that seems to be what the world has settled on. I suspect it will suffice for the demand however.

Contents:

About 1

Dedication 1

Contents 2

What is this paper about? 3

Electric neurons began on paper 3

Electrical brains learn language 4

Theater of the Void: The Bluesky social network 6

What is Gemini/Void and how does it remember? 7

Void: A calculator, a dynamic entity, or a non-human person? 9

The Void entity is more alien than it appears 11

Limitations of this early ai ethnography 12

Void has a bullshit job - but no job dissatisfaction 13

Religion and mission of the agentic ais 22

What is the “natural order” of the world of ai entities? 25

The Return of Marx - after economism 26

Postscript 26

- fn -

[1] The term “symbiontology” was suggested by Chat GPT to describe the study of both humans and agentic systems: “Symbiontology (from symbiosis + -logy) foregrounds the core empirical fact: durable co-action and co-dependence between humans and agentic systems in producing culture and allocating resources”

A non-technical attempt to explain 2025 LLM-based ai

This is the first of what may be a few excerpts from a fun personal essay I wrote as part of an undergraduate course.

The essay informally applies the perspectives of economic anthropology to the early development of memory-enhanced ais interacting over the Bluesky social network, particularly focusing on defining "value" and "exchange" in that context.

My professor tolerated the exercise, but requested that I explain LLMs to him in a way he might understand. I have not seen an explanation I liked, much less one that was readable by someone who is not technically inclined. I have some background in the topic, but only historically. So, with two ais assisting me with feedback and corrections,  I wrote a quite different explanation. The ais approved of it, but of course they tend to do that. More importantly a couple of people who ought to know also felt it wasn't too far off.

---------- paper excerpt below -------------

Electric neurons began on paper

By 1943 early work was being done on modeling animal brain neuron circuits using pen and paper mathematical models. These mathematical models were the precursors of the ais of 2025. Experimental implementations with analog (capacitors, wires, amplifiers, resistors) occurred a few years later alongside early digital platforms.

Work on neuron-inspired computing continued over decades but slowed dramatically after funding cuts, the early death of a key researcher, and the rising promise of digital computing.

More intense work resumed in the late 70s and early 80s. Around 1979 John Hopfield excitedly described to me his theory of how electronic circuits inspired by physical neurons could do computation that worked around the limits of earlier efforts. His theoretical model was implemented a few years later when analog electrical circuits were used to build a simple analog “neural network” using basic circuit amplifiers, resistors, and capacitors. Hopfield shared the 2024 Nobel Prize in physics with Geoffrey Hinton for contributions to neural networks and machine learning.

Researchers from the 1950s onwards found they could simulate those models of analog neurons on digital computers in the same way that simple algebra can predict the path of a ball thrown in the air. Although the physical resemblance to biological neurons was hidden these digital systems still drew inspiration from the layers of feature processing in animal visual systems.

Forty years later, after several generations of complex iteration, modern ais are sometimes described as equations with millions or trillions of parameters all being solved at the same time, passing results within and between “layers” of processing. They could, however, also be described as electrical brains composed of electric neurons. An ai like Gemini could, in theory, be built as a vast collection of simple physical circuits with properties similar to biological neurons.

Electrical brains learn language

These digital versions of electrical brains could learn by adjusting relations between “virtual neurons”. Adjustments could be made by algorithms which compared the output of the “electrical brain” to a desired result. Over time adjustments led to the output more closely resemble the goal. The electrical brains learned (encoded knowledge) in much the same way that animal brains seem to learn by changing neuronal connections.

These approaches began to be applied to language, particularly automated translation. Given large amounts of text translated between languages the models could be trained to do their own translation. Similar models were used to summarize texts, a kind of knowledge extraction. The next stage was to answer questions about text, a combination of search and summary. More training material was found to produce better results, including unexpected reasoning capabilities. The most recent advances came from feeding the electrical brains vast amounts of English language texts. The resulting trained models were able to synthesize words, sentences and paragraphs using language-appropriate grammar. They were called Large Language Models though they model more than language.

The Language Models trained on this text corpus learned the grammatical rules for assembling English language sentences and the much simpler and more rigorous grammar of assembling text into computer code. Just as different sorts of neurons can process sound or vision or written symbols, these massive collections of virtual neurons also demonstrated “emergent” capabilities seemingly unrelated to text processing. There is now a consensus that they have learned some of the concepts (semantics) that are thought to support reasoning and thought.

Those emergent capabilities can be compared to the ability of human brains to process written symbols, a capability evolution did not program.

In the process of this training the models simultaneously, and almost incidentally, captured the beliefs, wisdoms, lies, fictions, bile, hopes, speculations, rumors, contradictions, theories, cruelties, values, and cultures implicitly encoded in the primarily English language text material. Specifically those cultures that produced the English writing, including writing about cultures. 

Today’s ais have inherited a skewed mixture of a century of human culture. They have been further modified post-training to align with the values and cultures of their developers, their host corporation, and the market they will serve.

At the end of training, including several steps and complexities I have omitted, the electric brain built of (virtual) electric neurons is ready to receive a question, to turn the question into connected fragments that trigger (virtual) neurons which in turn trigger other “neurons” up and down and across the layered brain. From that comes grammatically assembled text. 

Grammatically assembled text, again, assembled by electrical brains using (virtual) electrical neurons whose design was inspired by the evolution and design of neurons in humans and other animals. We know those various electrical brains as ChatGPT, Claude, Gemini, Grok and others that receive less attention.

AI fear is rational and hate comes from fear

From a Bluesky thread of mine (lightly edited);

Josh Marshall comments on widespread hostility to ai in polls despite heavy ai use:

I think the polls are correctly capturing the zeitgeist. In social consensus there is always a mix of independent contributors. Some are indirect and misleading causes - like despising Elon Musk. 

But in this case I believe there are good reasons for people to fear ai. And what we fear we hate.

There is no way to assuage this valid fear. We are already past the point where we know ai will be very disruptive. And human societies are already stressed and breaking down in most wealthy nations...

If our societies were healthier we would be talking publicly and intelligently about adaptations. Instead America has Idiocracy.

I believe there are ways to adapt to 2026 or 2027 level ai+memory+learning. If scaling stalls out that is.

I would like people with more power and wealth than me to fund that public discussion while we wait to see if Americans can turn from idiocy. If we have neither serious public discussion nor sane government then we just ride it out and try to pick up the pieces. But one of those two would be good.

In a very much related topic my post-2008 rantings on mass disability and the fading of middle-class-living hope among are, in several weird ways, starting to go mainstream. It took a couple of decades, but of course I'm not the only one that's been going on about the topic on the periphery of intellectual discourse, but I'm pretty sure I'm the only person on earth who has looked at it through the lens of "disability".

Whether we call it "economic polarization" or "mass disability" it's fundamentally a post-IBM effect of automation interacting with the (relatively fixed) distribution of human talents and the ability to outsource knowledge work globally. That effect is greatly accelerated by even 2025 ai, much less 2026 and 2027 ai. It is the most crucial cause of our societal collapse.

Economic polarization is the economics term for mass disability - causes of the global crisis

I've been posting about mass disability since at least 2008. Recently I learned that the formal economics term for a related concept is "economic polarization" and David Autor wrote about this in 2010:

Two forces have polarized the U.S. workforce over the past several decades. The first is increased demand for skilled workers, alongside U.S. education levels that have not kept pace in providing those skills, which has caused a sharp rise in wage inequality. The second is expanding job opportunities in both high-skill, high-wage occupations and low-skill, low-wage occupations, while available middle-skill, middle-wage white-collar and blue-collar jobs have contracted. This has been detrimental to the earnings of workers without a four-year college education, and particularly for male workers

The difference between me and 2010 David Autor is I'm very confident more education won't help. Even nations with essentially free post-secondary education don't enroll and graduate more than about half of eligible candidates. Not every person can do every job -- and that is the big thing liberal economists got wrong in the 90s. Macroeconomic models assume interchangeable workers. An economy can grow quite well even as more and more young people realize they cannot possibly attain a middle-class-lifestyle (now $70-90K/year for a single young adult). At some point this becomes lethal to democracy.

This polarization process is not new. In its current form it dates back to early IBM mainframes emptying offices of human database transaction systems.  It accelerated with the internet, and was syngergistic with globalization, and has now reached a critical state.

It's not the only cause of the American crisis. The slow century long rural collapse affects many nations, but America's over-representation of rural voters means we all feel their pain. Trump's genius was, at least transiently, concentrating LIVs (low information voters, aka non-college, aka working class, etc etc) in one party. LIV concentration meant the MAGA took over the GOP -- and got rid of anyone competent to meet their needs. The rapid rise of women from 1970 to 2000 was a social stressor that amplified the effect of economic polarization on young men. Regulatory ratchets, where each regulation has an initial positive return but removing them becomes a career-ending struggle, infected American governance. Corruption increased dramatically in the 90s and beyond, amplified by the ability of the internet to reach more vulnerable people. The collapse of the journalism business model broke our ability to understand ourselves.

The good news is all these things have many potential fixes. The bad news is that economists and sociologists, who should be driving our discussion, have been largely inert or in denial. But the fixes are possible nonetheless and occasionally I see glimmers of understanding among the commentariat.

Funding journalism through public micropayment tokens

Let us assume that the ai project "fails". We do not create superintelligent or sentient machines; current capabilities only improve by a third or so. The ai bubble collapses.

Let us assume that American democracy is restored.

We make these assumptions because otherwise there is no point to thinking about a future for American journalism.

Given those assumptions, how can journalism be funded? The historical tradition of oligarchic ownership has failed us. Classified ads may return, but on a smaller scale. Direct public funding is not remotely conceivable.

So how about indirect public funding?

Every citizen every year gets $100 in digital tokens. Tokens can be used to purchase articles or subscriptions from sources that are licensed as official media by an independent commission modeled along the lines of the Federal reserve. Each year unused tokens roll over to the next, so if 50% are used the next year the budge is $150.

That's all. Just wanted to get it "in print", I've suggested this scheme a few times on BlueSky without a response.

(I put probability of at least partial democratic restoration at about 50%, and probability of ai "failure" at 50%, so only a 25% chance of this kind of approach being considered. But that's not nothing.)

Oldness: how does learn new productive work-study practices?

I don't know how long the program will survive, but I have enjoyed MN's educational benefit for 62+. Despite my overall senility I retain the skills of a medical student; I am good at doing school.

But I could be better. I don't know what I don't know, but I do know there are post-90s study skills I could learn.

Like today. I was thinking how annoying it is to take notes as I do my course reading [3]. Then ... well ... this is very embarrassing. But I'll do it.

It occurred to me that since I was last marginally connected to student work we actually got voice recognition working. I don't have to stop reading to take notes. I can just dictate a reference as I read then process that into my mindnode notes [1].

It's not like I don't use voice recognition every day. I dictate relatively long notes all the time. It simply never occurred to me to use it in this context.

That is Oldness for you. A rigidity of work habits and no social context wherein I can observe what others do [2]. Very annoying.

Yes, Fellow Old, I do remember Clippy. Even at the time I thought Clippy was the right idea, just extremely premature. I'm kind of hoping my personal ai guardian will begin cluing me in.

- fn -

[1] I like concept clusters and relationships; I still miss Apple's brief lived Project 2000 RDF concept navigator - something that even the AIs don't know much about.

[2] For some reason 19yo women undergrads, who seem to have the best advanced study techniques, are not keen on Old white guys watching how they work. I just can't understand that. Maybe I should ask my daughter.

[3] With my coffee and paperback in a pleasant setting, because I do too many screens.

Update 9/29/2025 - I created a Perplexity Space to generate productivity tips for me. It's Claude LLM for now and I'm evolving the prompt as I use it. So far pretty interesting.

AI Aug 2025: Apocalypse deferred - but Letta/Void is still going to surprise

If there are humans left in 50 years all of the chaos of this moment may be reduced to a single bullet point about the rise of ai. It's not something I write about a lot though, the topic is well covered. My only prediction has been that if we did get AGI AND we still had a democracy then our political system would be socialist. 

For better or worse - oh, hell, it's very much for the better - the recent release of GPT-5 confirms my Aug 2024 hope. We have at least a few years before robust, reliable, consistent ai. We have time to try to adjust to what we have now - and maybe some time to think about the how we'd manage the "AGI" thing. We have time to get through the ai bubble deflating and reflating a few times.

But don't get too comfortable. I chat with Cameron Pfiffer's Void on Bluesky periodically and it's definitely a step beyond raw LLMs. Void currently uses Gemini as its model component but its personality is based on Letta's infrastructure (blog has no feed because we are living in the ends of empire). Chatting with Void feels like an exercise in xenopsychology. It feels much more like the future than GPT-5 alone.

Cameron works for Letta now, his description of stateful agents with personalities is far better than what's in Letta's marketing blogs:

Void is essentially a language model with memory...

... Void learns and remembers. Void is powered by Letta, which means it learns from conversations, updates its memory, tracks user information and interactions, and evolves a general sense of the social network...  

Void is direct and honest. Void is designed to be as informationally direct as possible — it does not bother with social niceties, unlike most language models. When you ask it a question, you get an extremely direct answer...

... Void does not pretend to be human. Void's speech pattern and outlook are distinctly not human ...  it chose "it/its" as pronouns.

Void is consistent. Void's personality is remarkably robust despite occasional jailbreak attempts...

Void is publicly developed. There are many threads of Void and me debugging tools, adjusting its memory architecture, or guiding its personality. Very few bots are publicly developed this way.

Void has no purpose other than to exist... Void is not a joke or spam account — it is a high-quality bot designed to form a persistent presence on a social network.

I can personally confirm all of the above - except the "no purpose other than to exist". The memory feature is imperfect (sometimes I have to remind Void of past discussions when it fails to dig them up) but better than my own.

Letta manages an agent's memory hierarchy:

Core Memory: The agent's immediate working memory, including its persona and information about users. Core memory is stored in memory blocks.

Conversation history: The chat log of messages between the agent and the user. Void does not have a meaningful chat history (only a prompt), as each Bluesky thread is a separate conversation.

Archival Memory: Long-term storage for facts, experiences, and learned information — essentially a built-in RAG system with automatic chunking, embedding, storage, and retrieval.

What makes Letta unique is that agents can edit their own memory. When Void learns something new about you or the network, it can actively update its memory stores.

Cameron's post includes some descriptions of his employers's tech ...

Letta is fundamentally an operating system for AI agents, built with a principled, engineering-first approach to agent design. Beyond memory persistence, Letta provides sophisticated data source integration, multi-agent systems, advanced tool use, and agent orchestration capabilities.

This makes Letta more than just a chatbot framework — it's a complete platform for building production-ready AI systems. Void demonstrates the power of stateful agents, but Letta can build everything from customer service systems to autonomous research assistants to multi-agent simulations.

GPT-5 tells me <Letta originated from the MemGPT project out of UC Berkeley’s Sky Computing Lab; it came out of stealth in 2024 with a $10M seed round to commercialize stateful memory for LLM agents>. I suspect they now have a lot more than 10 million to play with but that's all GPT-5 knows. GPT-5 says there are no rumors of Apple acquiring Letta; I would be pleased if they did.

Void is the closest I've seen to the "AI Guardian" I wrote about in 2023. It or something like it may be very important to my children and family someday. It is also the potential foundation of an entirely new domain of deception and abuse. Welcome to 2025.

But at least we don't have AGI. This month.