A few links-and-comments from the week:

(1) Don’t Believe Them

Ezra Klein interviewed Jasmine Sun this week about the data center backlash. It’s an interesting conversation, well worth a listen.

One bit that stood out to me was this point about data center activists simply saying “I don’t believe them” in response to promises from the hyperscalers.

The bit that I want to emphasize is: Good. You absolutely should not believe the data center proponents. Every assurance they make is subject to subsequent revision. It’s 2026, and we should all know this by now.

I recently read Eric Ries’s new book, Incorruptible: How Good Companies Go Bad and How Great Companies Stay Great. It was delightfully much more radical than I expected. (I don’t usually see startup gurus take a sledgehammer to Milton Friedman’s corpse. More of this, please!)

Ries’s main thesis is that well-meaning companies have to take governance and corporate structure very seriously, or else they will get sabotaged by private equity/hostile board members/other avatars of the worst excesses of present-day capitalism. Good companies go bad because we prioritize immediate shareholder value over all other types of value-creation, with disastrous consequences both for the companies themselves and for society writ large. (I’d recommend reading Megan Greenwell’s Bad Company and Catherine Bracy’s World Eaters first, but Ries’s book is a welcome addition to the genre.)

Anyway, as Ries makes clear, the question that you have to ask whenever a tech company (or any large company) strikes a deal with a local government is “what penalty does the company pay if they go back on the deal?” If the penalty for reneging is cheaper than the cost of fulfilling the stated terms, then you should assume that the company will renege. (Because that’s how the company maximizes shareholder value in the upcoming quarter. Everything else is secondary.)

We’ve seen this phenomenon for decades with environmental/health and safety regulations. It is cheaper for coal companies to pay the fines than to follow the law, so the coal companies just pay the fines. We’ve also seen it in Elon Musk’s legal arbitrage. We know how this goes.

When a company says they are “going to treat the water with chemicals so it doesn’t flow into the lakes,” that is a limited-duration promise. What the company actually means — the only thing that it can mean, in fact — is that it is willing to make this promise to the community right now in order to sign the deal it is pursuing. At some later date, some C-suite executive (or outside consultant, or politically-connected VC, take your pick) is free to notice that treating the water with chemicals costs a lot of money, and that it would be cheaper to just stop treating the water with chemicals. And then, unless the enforcement mechanism is absolutely ruinous (and, in 2026, they never are), the company will abandon its former promises. And probably give the exec a bonus.

The companies could build out the electric grid, constructing massive solar farms to support their massive data centers. But it would be cheaper (for them, in the short term) if they didn’t. So, in practice, they generally won’t.

You don’t have to believe the hyperscalers are particularly villainous companies to conclude you shouldn’t believe their promises. You just have to believe they behave like large companies, and obey the same short-term profit-maximizing incentives as every other large company.

(2) The Situation with Situational Awareness 

The whole Leopold Aschenbrenner thing is such a fascinating trainwreck. If you haven’t followed the details, read Liz Lopatto’s (perfect) writeup from last week.

The TL;DR version: Aschenbrenner (a former OpenAI and FTX employee) wrote a manifesto in 2024, insisting that scaling laws would go on forever and superintelligent AI would completely transform the economy by 2027. It included a bunch of chart-crimes. Silicon Valley loved it. Then he launched a hedge fund. The manifesto was great advertising, and the investment thesis was basically “leveraged bets on AI lines going up.” Good thesis, for the past couple years at least. He made a ton of money for his investors. But then, last week, he lost basically all of it and had to sell the assets to mega-investor Ken Griffin.

Aschenbrenner’s chart-crime. This is bad. Do not ever make a chart like this.

This seems to be less a sign of that the AI financial crash has finally arrived and more a sign that sophisticated AI thinkers don’t know what they don’t know about finance and hedge funds. (As always, read Matt Levine).

Also, too: this all happened on Thursday last week. Last week, the news was that Leopold Aschenbrenner had lost roughly $35 billion of his investors’ money. And then this Tuesday, he relaunched the fund with a fresh $400 million of private investment. AND OVER THE WEEKEND THE GUY GOT MARRIED!

I am so, so curious about the vibes at that wedding. Congratulations to the happy couple. May the rest of your lives be so much less hectic than the beginning of your marriage.

That’s all for this week. August is going to be a pretty light month, newsletter-wise. I’ve got some vacation planned, and am training for a marathon right now (in August. In DC. I may not have fully thought this through), so I’m basically dividing my time between (1) running and (2) being exhausted from running.

One way to know the data center backlash is working is that the New York Times opinion section is freaking out.

Over the weekend, the Times published a guest essay titled “The Environmental Case Against Data Centers Is Misguided,” by Experimental Psychologist/Substacker Adam Mastroianni. The argument he makes is a good indicator of the current industry talking points. To wit:

-People sure do seem to hate data centers! But should they, though? When you think about it, data centers really aren’t so bad.

-Here are some statistics about energy usage and environmental impacts. They were generated by a handful of industry-funded bloggers and think tanks. Surely their modeling assumptions are pure and good. And, look at that! They use practically no energy at all. It turns out that data centers are way less harmful than golf courses. You don’t see people protesting golf courses, do ya?

-Anyway this is just a new form of NIMBYism. All those data centers are bound to get built somewhere. So, when you think about it, powerful communities stopping data centers in one place just means they’ll get built in poorer communities. Surely there are better ways to fight the AI industry.

-And hey, I’m critical of the AI industry too! There are good and bad things about the AI revolution. But what we should really be focused on is responsible use. Everyone, individually, ought to think about the right and wrong ways to incorporate AI into our lives. That would be far more effective than the data center backlash. Let’s engage in this totally different conversation about ethics and personal responsibility.

I’m seeing this same argument pop up a lot, particularly in techno-optimist/reactionary-centrist circles. And there are just three quick things I want to say about it:

1. The math on energy and water usage is complicated. (Read Matteo Wong, “The Truth about AI’s Water Use” and Alex Reisner, “Generative AI Is an Engineering Disaster.”) It’s complicated because an awful lot depends on design choices and cost calculation. You could build a data center that uses basically no water, and that is powered entirely by solar and batteries. But that would cost a lot more money and take a lot more time. The hyperscalers are treating AI as a race, and are burning cash at a phenomenal rate. So of course the actual environmental costs are much higher than the ones calculated by the industry’s bloggers-of-choice.

2. Here’s something that isn’t complicated: the sheer, unprecedented scale of their proposed buildout. The hyperscalers are aiming to build 10 gigawatt data centers. 10 gigawatts of power is around the same energy consumption as New York City during a heat wave. That’s… a lot.

2a. Anytime you have an emerging industry, no established industry standards, and big questions about its impacts, some of the biggest fights will be second-order arguments over what we should measure and how we should measure it. Industry-friendly blogger Andy Masley has made a name for himself by “debunking” environmentalist claims. He mostly does so by changing the scope of analysis to things like the cost of a single ChatGPT prompt. It’s a fundamentally silly argument to get involved in. The AI executives are saying (basically) “we plan to consume so much power that we will soon need to blot out the sun with space data centers,” and their favored bloggers are then chiming in to say “but when you think about it, that’s not much energy usage at all!” Massive new data centers are, in fact massive. This isn’t complicated. The people saying otherwise are either grifters or marks.

3. And, also, this is entirely about political power. Of course it is. They wouldn’t be complaining in the NYT if it wasn’t working. Last week I was listening to David Roberts’s Volts podcast last week, and he noted that the AI industry has made a colossal blunder by not just, like, paying for goodwill from all these towns. It would be relatively cheap, compared to what they’re spending, to just build schools and libraries and clean energy grids for these towns. But of course OpenAI and Elon’s company and all their competitors aren’t actually doing that. It would cost them money and time, plus they would have to explain to their next round of investors why a chunk of their money went to building a library in Ohio instead of making the Singularity happen any minute now. They could make these communities better, just like they could use the data center buildout as a tool for rebuilding the energy grid and radically increasing renewable energy capacity. But they won’t, because that isn’t the game they are playing. It never has been, and it never will be, unless we completely shift the power dynamics.

I wrote a couple months ago about data center activism as a bankshot. The reason why I think this activism is very good, and very promising, is that it has the potential to interfere with the AI financial bubble. If that bubble pops, it unsettles the power dynamics shaping the AI future. If you think it is bad that Sam Altman, Elon Musk, and a handful of their billionaire buddies are effectively unconstrained, then the most effective thing you can do is pick a fight over data center siting in your area.

Mastroianni concludes by arguing for a politics of personal responsibility. He writes:

I’m going to presume this argument is made in good faith. I don’t know Mastroianni. Maybe he really is deeply critical of the current trajectory of A.I., and also deeply concerned that the A.I. critics are going about things the wrong way.

But let me just say that his theory-of-change absolutely sucks. The A.I. industry is powerful and fragile. A.I. opponents have landed on a strategy that successfully generates collective action and challenges the industry where it is most vulnerable. And this guy thinks what everyone ought to do instead is focus on their own personal decisions and vague social norms.

The tactic is working. Bad, powerful people are freaked out. Don’t listen to their favorite bloggers who say “well, when you think about it from just the right angle, data centers aren’t any worse than golf courses.”

When you have a tactic that is working, and your opponent doesn’t have a good countertactic, the smart move is to keep doing it.

One of my long-running arguments on this blog (and in the forthcoming book!) is that Silicon Valley in general, and the AI industry in particular, runs on futurity.

Futurity is the ambient, collective sense of what sort of technologies and business look and feel like the future. It’s just vibes, but with trillions of dollars on the line. The futurity side of Silicon Valley is much bigger than the engineering side of Silicon Valley. SpaceX and Palantir’s market caps aren’t dictated by their product lines or profit margins; they're determined by the companies’ ability to seem like they’re building the next-big-thing.

Futurity is a fickle thing. You have it until you don’t. (The metaverse was the future, and then it very much was not.) All of the big AI players recognize this. So they work very hard to keep up appearances that (1) the AI revolution will be bigger than the industrial revolution, and (2) the AI revolution is coming so very very soon.

This takes the form of a communications campaign. The purpose of the campaign is quite simple: whenever it starts to feel like progress in AI is slowing down, there needs to be another announcement or product demo that makes it feel like the future is fast-arriving.

(The comms campaign, mind you, is separate from the engineering effort. So long as you maintain the aura of futurity, your engineers get limitless resources to build impressive things. It can both be the case that Claude Fable has far more capabilities than last year’s top-of-the-line models, and that wholesale transformation of the economy remains a fairy tale for investors and discussion board maniacs.)

The AI industry’s comms campaign isn’t particularly complicated. All they do is maintain a drumbeat of gee-whiz stories. If a couple months go by without a public conversation about the radical pace of AI change, then you manufacture one through the publication of another online manifesto or a sign-on letter.

Which brings me to Monday’s AI news. Here’s the NYTimes headline: “Nearly 200 Economists and Tech Leaders Warn of A.I. Threats.” It’s a story about a new sign-on letter, titled “We Must Act Now: A Statement on AI’s Transformation of the Economy.” Some very big names signed on to this letter, including 16 Nobel laureates. Sounds pretty newsworthy, no?

Except, when you read the letter itself, it is just about the most anodyne document imaginable. Here is the text, in full:

That’s… nothing at all. AI might become radically more powerful over the next decade (sure). This could be a big deal for the economy (uh huh, yup yup). So economists, policymakers and tech leaders should do a lot of thinking about this stuff. (wait. What?)

This isn’t a call-to-arms. It’s a request for research funding.

In 2023, the Future of Life Institute published the AI Pause Letter, demanding a 6-month moratorium on the development of new foundation models so that policymakers could catch up with the radical pace of AI development. There was no AI pause. Current models are far more powerful than GPT4, and we are not noticeably closer either to the AI jobs apocalypse or the posthuman future.

In 2025, the AI 2027 scenario predicted that the AI future was barreling towards us, and that by 2027 (next year!) Superintelligent AI would likely doom all of humanity in the absence of extremely competent diplomatic coordination among international governments. I could write a long post about how none of this is actually happening the way they predicted, but that would be beside the point. The strategic purpose of the scenario was to contribute to the steady cadence of AI futurity hype. The scenario did what it was intended to do.

By comparison, this economist letter is remarkably anodyne. What academic economist would disagree that (1) AI might keep improving, (2) this might have economic impacts, and (3) economists and policymakers ought to devote some quality brainpower to this sort of thing?

(The equivalent in my field, political science, would be a mass sign-on letter saying (1) American elections might become more polarized and less competitive over the next decade, (2) this could have implications for democracy, so (3) governments should fund a lot of political science research and listen to the recommendations of political scientists. Practically all political scientists would agree that there should be more research funding for political science, and that policymakers ought to listen to us.)

From a strategic comms perspective, the point of this letter isn’t the substance of the letter. It is substance-free. The point of the letter is to combine the title (We Must Act Now!) with the signatories (A bunch of Nobel laureates, several of whom are skeptics). Because that’s how you maintain the cadence of public conversations about the AI future.

It sure would be nice if mainstream journalism outlets stopped falling for it.

My second book, Analytic Activism: Digital Listening and the New Political Strategy, was published ten years ago. It has been a very long decade.

One of my projects this summer is to revisit the book and do some serious thinking about what has changed in the intervening years. I did the same thing with my first book, The MoveOn Effect, back in 2022.

I reread the book this week. I now have a large pile of notes and just a handful of authorial regrets. Today I just want to set the stage a bit for the essays to come.

Let’s start with the regrets. Analytic Activism is a good book. It could have been a better one.

The research for this book spanned the years 2013-2015. It is very much a late-Obama-era book. It is also a book from what we now know was the peak of digital media’s platform era. It was written at a time when online petitions were very popular and (in certain contexts) quite effective. It was a time when the ascendance of Facebook and Twitter was more opportunity than problem.

There is, as a result, a certain quaintness to the book. Reading the first few chapters feels a bit like reading about the politics of Pompeii, just prior to the volcano. It’s not that I was wrong about the politics of petition platforms — it’s just that those politics were about to be rendered irrelevant by much larger social forces.

Chapters three (digital petitions) and four (Upworthy/platform virality) are simultaneously the book’s strongest chapters and its most dated chapters. I handed in the full revised manuscript in March 2016. Back then, it still seemed like someone would beat Donald Trump in the Republican primary. The book was published in December 2016, just a month after he was elected. The middle chapters of the book were, in a sense, immediately outdated. If you want to know what reformist/progressive digital politics was like in 2013-2014, these chapters are an excellent time capsule. But if you want to understand digital politics over the decade that follows, they offer far less relevant insight.

Another regret is the title. In all my early files, I called it “Analytics-based activism.” That seemed unwieldy, so I went with “Analytic activism” instead. I wish I hadn’t. The book isn’t delineating a distinct form of activism that is more analytical than the rest. The book describes how netroots activist organizations make use of data analytics (digital listening) and testing to develop new tactics and strategies. The title was a misstep.

Rereading the book, I can quite tell the stage of my career in which it was written. It’s a second book, but also a pre-tenure book. It is very much a sequel to The MoveOn Effect, revisiting the same set of organizations and digging deeper into how they operate. It tries to simultaneously speak to political practitioners and the research community. And that’s fine, but it also bogs down the writing. At times it feels like every alternating paragraph is meant for a different audience, delivered in a different register. I can tell, in hindsight, that I hadn’t yet worked out what kind of writer I wanted to be.¹

But those are all just notes for my younger self.² The real point of this rereading exercise is to generate some new ideas — either about things I got wrong then, or unexpected developments over the intervening decade. I have five such ideas, each of which will become an essay in this series:

(1) Revisiting and updating the media theory of movement power. This was the big idea in chapter 1 of the book. It probably should have been the central concept running through the entire book. And I think it still holds up quite well. I’d like to think through how both the political change and the media change of the past decade affects what political activists do and where/when they succeed.

(2) What do we talk about when we talk about (digital) organizing? This is going to be more of a historically-focused essay, looking at the differences between how Saul Alinsky and Ralph Nader approached power-building in the 20th century. The Naderite approach has dominated among reformist-progressive U.S. organizations for several decades, which (I believe) has major implications for how present-day organizations have approached digital tools, listening, and tactics.

(3) Under what circumstances do people in fact prefer empowerment to convenience. One of the bits from chapter 5 that I found most interesting in retrospect was the AFL-CIOs experiment in “governance gamification.” This never made it past the pilot stage. I think it’s worth thinking through the reasons why.

(4) Does generative AI ruin the entire tactical repertoire of analytic activism, or just most of it? The most powerful activist organizations circa 2013 were generating big online numbers at a fast pace. Generative AI has made it easy to fake a lot of that. I want to spend some time thinking about what that means for tactical repertoires. (This essay is probably going to be the big bummer of the series.)

(5) Research note: Organzations are assemblages. It’s really-damn-hard to keep the whole thing in view. This last one will be pretty much entirely for the research community. Both Analytic Activism and The MoveOn Effect took a particular approach to how we study social movements/political advocacy/civil society/political engagement. I have long advocated for focusing on the “organizational layer” of politics. It’s been a decade, and I find I have a bit more to say on the subject.

Each of these essays will be a significant undertaking. My hunch is that I’ll write one per month, mixed in with the usual essays on technology, politics, and power. I’m looking forward to writing them. Let’s see how this goes…

I stopped by the Great American State Fair on Tuesday. The website listed the day’s theme as "Innovation, Technology & Progress.” It doesn’t get more on-theme for The Future, Now & Then than that.

I documented the trip in a bluesky thread, the highlights of which are featured on Flaming Hydra (have you subscribed to Flaming Hydra yet? It is the absolute best. Please subscribe.).

The most striking part of the experience, really, was just how boring it all was. I did not expect the Fair to be good, but I expected it would be an interesting sort of bad. Trumpist aesthetics include an awful lot of trolling and demagoguery. Both are bad and socially corrosive; neither are boring. The UFC fight on the Whitehouse lawn was an absolute assault on the senses. The Fair was something else.

I have walked the entire mall and seen everything.

The overarching vibe is boredom. I cannot emphasize enough just how little there is to DO here.

Imagine a state fair with one Ferris Wheel & no other rides. Three event stages, but no performance schedule. Dozens of booths, with nothing inside.

— Dave Karpf (@davekarpf.bsky.social)
11:01 PM • Jun 30, 2026

The Great American State Fair isn’t like a trip to Epcot, or to the World’s Fair. There just isn’t nearly enough to do there. I have spent 14 hours at Epcot with small children. I do not recommend it. But the kids never lacked for entertainment at Epcot.

It also isn’t much like a state or county fair. It has the trappings of one. There’s a Ferris Wheel. There is a daily rodeo. There are (air conditioned) booths, where you can learn about states or federal agencies. But it’s all half-done, as if the planning team got DOGE’d early and no one noticed the to-do list left undone.

Really, what are we doing here, folks?

It also — and this is the real, inexcusable flaw — is not like the Smithsonian. The National Mall is home to world-class, free museums. If you are going to hold a 250th anniversary celebration on the National Mall, it either has to be distinctly different than the Smithsonian, or else it has to be better than the Smithsonian.

The state exhibits are lackluster. But they’re particularly an embarrassment because the National Museum of American History is right there, across the street, just steps beyond the fencing. The largest display belonged to SpaceX. It was underwhelming in its own right, and rendered downright pathetic when you consider the National Air & Space Museum is two blocks away, open to the public, with much better exhibits.

The crowds are sparse, on par with the tourist traffic you’d see on the Mall during a normal summer day. The Fair is much too spread out, extending across seven city blocks. It runs eleven hours a day, for two weeks straight. And, again, there is nothing to do there.

When I visited, the largest crowd I found was watching the rodeo. That crowd filled a pair of small bleacher seats. Imagine the away crowd at a high school football game, in a state that doesn’t particularly care about high school football. That’s as big as it got.

I walked the entire Mall and saw everything there was to see in under two hours. It is punishingly hot outside. People aren’t going to spend the day just hanging out at the fairgrounds. And they aren’t going to cluster around the big events or attractions, because they either don’t exist or haven’t been publicized.

Another reason for the lack of crowds is the complete absence of any public schedule of events. The Fair has three performance stages. You cannot search online to find out who is performing when. You are not handed a program upon arrival. It turns out that when you register (for free, no actual ticket required), you are added to an email distribution list. Each morning, they send out the day’s schedule of events. I signed up on Tuesday afternoon, so I didn’t receive an event schedule until Wednesday. (Glenn Beck appeared on the Wednesday mainstage. His was the only name in the lineup that I recognized.)

This is no way to run a fair or a festival. It’s no way to run an event of any sort.

The saddest moment of my visit came while watching a performer on the National Endowment for the Arts stage. He had an audience of one. In between songs, he thanked the crowd and asked them (her) to follow him on Instagram. Logistically, this is unavoidable. People will not show up to a concert if no one informs them it is happening. Particularly not in 100 degree heat.

The worst part of the Fair is the Freedom Truck. DO NOT ENTER THE FREEDOM TRUCK! Inside you will find a Prager U exhibit on American history. Again, let me remind you that the National Museum of American History is RIGHT ACROSS THE GODDAMN STREET, PEOPLE!

The second worst part of the Fair was everything else.

animated George Washington, powered by Prager U. This broke me. I am broken now.

My hunch is that the event organizers got spooked by all the cancellations, so they decided to keep the performance schedule quiet. Another possibility is that they are still scrambling to fill performance slots, and they simply don’t have the schedule locked in until the day-of.

I cannot decide which possibility is more embarrassing. These are not professionals. They are not qualified to plan a toddler’s birthday party. I cannot imagine being this bad at the nuts-and-bolts of your job, in public, where everyone can see you being see your basic incompetence on full display.

Just imagine planning a cross-country trip to attend this fair: It spans two weeks. There is no advance agenda to plan around. So you… pick a weekday at random? You arrive to find empty exhibit halls and overpriced food. I hope you like Glenn Beck, and I hope you happened to show up on Glenn Beck day? The nice thing, I suppose, is that you can get real close to the stage. But that’s because attendance is so low and spread out that you feel basically alone.

Meanwhile, the Smithsonian museums are just past the fencing. You could have a better experience, cheaper meals, and (much) better air-conditioning just a block or so away. But it isn’t a special occasion for the Smithsonian. It’s just a weekday. This is what they do all the time.

So it isn’t a state fair, and it isn’t the World’s Fair. The in-person experience is like the empty shell of a fairground experience. One Ferris Wheel, and no other rides. Three event stages, but no performance schedule. Dozens of booths, with nothing to offer visitors.

But what it does have is cameras, all pointed at those event stages. The performances have the texture of a Congressman giving a fiery speech to an empty chamber, broadcast live on C-SPAN, for an audience of no one in particular.

The Great American State Fair isn’t meant for the attendees at all. It’s online content. It’s meant for the streamers and online viewers.

That’s what it most reminds me of: it’s a digital Potemkin Village. They didn’t bother to create a real fair, or bring people together to celebrate the country’s 250th anniversary. They just threw something together that would look good-enough for the propaganda outlets, and called it a day.

And that’s the element of Trumpism that it most invokes: the rank amateurism… the slapdash construction… the sense that they just can’t be bothered to try anymore. Planning a national, multi-week event is a huge undertaking. It requires planning and execution. It is hard work. And what visiting the Great American State Fair reminds you of is that our current government just doesn’t do the hard work any longer.

Today marks the five year anniversary of The Future, Now and Then. This calls for a navel-gazing post, and maybe a slice of cake.

Photo by Will Echols on Unsplash

The title, “The Future, Now and Then,” was initially a bit of self-effacing humor. I figured I would be writing about tech futurism (the future, now), my WIRED research (the future, then), and contemporary political commentary (now). And I expected I’d only be writing occasionally (now and then).

The writing was indeed pretty occasional during that first year. 2021-2022 was also a low point for me as a writer. I had outlines for a dozen or so essays, all of them potential building blocks for the book. I was making progress on none of them. It just didn’t seem like I could get anything on the page. In July 2022, I got real mad at myself about a couple missed deadlines, and decided that the new rule was that I would write something every week.

That… worked. I’ve written 200+ posts since then. The Future, Now and Then has over 10,000 readers. The book is coming out in January. It’s the best thing I’ve ever written. I’m certain it wouldn’t be nearly as good if I hadn’t had this platform to work through the ideas in public first. I think I’m slowly starting to get the hang of this stuff.

A few stray observations:

• Five years in, I still cannot decide if this publication is a blog or a newsletter. The genre boundaries are fuzzy, and it usually doesn’t matter. Mostly I think of it as a blog-with-an-email-reader-list. But I’ve written a lot of drafts over the years that I would’ve published to a blog, because they didn’t merit clogging peoples’ inboxes. The closest analogue is a weekly opinion column. Once a week, I vent whatever research observations or public commentary have been rattling around my head. Sometimes that’s a book review or a big-picture essay, sometimes it’s a hot take or a rant.

• The fuzzy genre boundaries have exacerbated my oddball-academic status. Some of my academic research colleagues subscribe to this blog/newsletter. I show up in their inbox every week, and a subset of my posts are topical for their interests. They think I’m fantastically productive. But the vast majority of my field don’t read this column and many of them… kind of figure I’ve stopped publishing entirely? (Can’t blame them. My Google Scholar footprint is atrocious.)

• An interesting byproduct is, when I attend conferences and workshops, people either really know my work or they really don't. Those who don't are a little suspicious. ("Oh, you have a newsletter? That's... nice.”) Which, y'know, fair. It feels like a smaller version of the bretbug thing. I learned pretty quickly that the Bret Stephens episode had made me Twitter-famous, which is very much not the same thing as internet-famous or actual-famous. People who are extremely online will still drop references. Everyone else doesn't know what I'm referring to, and is a little unsure why they should care. (Which, again, fair! It is extremely niche and mostly pointless.)

• I don’t much bother with peer reviewed articles anymore. Maybe I should, I honestly can’t decide. For the types of intellectual contribution I want to make, I find the peer review process to be too slow and cumbersome to bother. I have deeply mixed feelings about that — I think peer review is good, and important, and mostly not-for-me. It would probably be better if I set aside the time to turn some of my longer-form essays into journal submissions. (Could “On Political Communication as a Mission Science” have been a journal article? Yeah, I think so. Would it have been worth adding ~2,000 words to the text, spending six months on minor revisions, and waiting a year+ for it to be published? Eh, I’m not so sure.)

• Geez, there are a lot of you. Ten thousand readers is a much larger audience than I ever would have planned for. It is very cool and more than a little humbling. (So, y’know, thanks! Have some cake.)

• And/but… 2021 was basically Peak Substack. Substack was the future of journalism, and the social media ecosystem was primed for sharing these newsletter-essays. That time has passed.

• Which brings me to the Substack-Beehiiv transition, and some more substantive points about the state of online publishing right now. The chart below is my subscriber growth over the lifetime of the publication on Substack. Loosely speaking, the publication added ~500 readers in its first, sporadic year (June 2021-’22), 3,600 readers in its second year (June ‘22-’23), 3,000 readers in its third year (June ‘23-’24), 3,500 in its fourth year (June ‘24-’25), and only 700 readers in its fifth year (June ‘25-26). The growth rate slowed around March ‘25. Either this publication hit a ceiling, or Substack’s discovery engine stopped working so well for free publications like mine.

The declining growth rate was one factor in my move to beehiiv. Substack has invested heavily in discovery features that help to juice subscriber growth within its newsletter ecosystem. I knew moving to an alternate platform would decrease the reach of my writing, but also noted that the list growth was already trending down.

So far, after three months on Beehiiv, I’ve found that people just aren’t signing up anymore. Every time I would publish a Substack essay, around 20 people would unsubscribe and (for the most part) more than 20 people would replace them. At Beehiiv, around 20 people unsubscribe every time I post, and 2-6 people replace them. My dormant Substack now has almost 500 more subscribers than my active Beehiiv.

I’m glad I moved off Substack. It was the right move, and the right time. And everything else about Beehiiv has gone smoothly. Open rates are steady. I’m no longer routinely embarrassed by the platform I publish on. Also, importantly, several dozen of you have stepped up as paying subscribers, which has been enough to cover the publishing costs. (For you, corner-pieces of cake!)

But oof, yeah, slowly leaking readership is a bummer.

There are a few potential explanations. It’s possible that I have basically reached the ceiling for this sort of writing. Ten thousand people is a lot! I write in a particular, obscure niche. Also, the old Substack probably appears in search above the newer beehiiv. Discoverability might improve over time.

My hunch is that this is part of a broader trend: the open web is dying. Chatbots and AI search are replacing it, and they absolutely suck for independent publications. I suspect that’s the main story here. (also Bluesky, I love you, never change, but you really don’t drive new email signups.)

A few years ago, Max Read wrote a great piece about newsletter/blog audience growth: “Matt Yglesias and the secret of blogging.”

I’ve kept Read’s rule of thumb in mind for years now. It’s simple and comforting: Write what you want to write. Write regularly. Readers don’t punish you for straying from your lane. The basic reason I have a lot of online readers now is that I have spent a lot of years doing a lot of online writing, and that’s how the online writing ecosystem functions. Don’t sweat the numbers; just keep writing.

I find it interesting and noteworthy that, at least for The Future, Now and Then, this is no longer true. Around 20 people will unsubscribe after receiving this post. 2-3 will subscribe.

And look, this is all low-stakes. I don’t write this blog/newsletter in order to build an audience — I write it because I’m a complete sicko who derives joy from the act of writing — but it does set a mood that I have to work to combat. It has the feel of a standup comic, performing while a small stream of people head for the exits. And it leaves me pondering the broader decline of the open web.

Looking forward, I have some ambitions for the months and years ahead. I’m going to keep writing (obviously). I’m planning an essay series reflecting on my 2016 book, Analytic Activism, 10 years later. I’m also planning to write on AI and activism. And I have a series of essays in mind about the past, present, and future of the climate/environmental movement.

This fall, I’ll switch into book-hype mode. There are some stray bits from the Wired project that didn’t make it into the book, so I’ll publish those as blog posts here. I’ll also keep writing cranky book reviews, and keep howling into the void as the whole country falls apart around me.

Basically the way I see it, now that I’ve written my way through one book, is that I’ll keep using this platform to hash out the main ideas for the next one.

Thanks, as always, for reading. If nothing else, I can say with certainty that writing this publication has been good for me.

Here’s to another five years. Have some cake.

Three medium-sized thoughts to share today:

I teach a class every spring semester called “History and the Digital Future.” This is the class that came out of the WIRED project. We read old magazine articles, discuss what the digital revolution looked like back then, and see what lessons we can glean from the imaginary futures that never arrived.

It is my favorite class to teach. I figure out something new every time.

The ah-hah moment this past semester came while we were discussing Kevin Kelly’s 2005 essay, “We Are the Web.” This is actually my favorite Kevin Kelly piece (mostly because he spends time reflecting on stuff he got wrong in the 1990s, and I’m an absolute sucker for self-critical retrospectives). We Are the Web examines three time periods: 1995, 2005, and an imagined 2015. It is a classic Web 2.0 essay, positively brimming with the excitement about online community participation that defined internet culture in that era.

Here’s how Kelly imagines the future, circa 2015:

There’s a nice puzzle to unwind here. Reading that first paragraph, it strikes me that Kelly is at least a little bit right. In 2026, we kind of do live in a world where, on average, everyone is more-or-less an online writer of sorts. If you count Twitter/X and Instagram and TikTok, we are absolutely awash in content. This is the world of social media streams, and it is a different world, a different culture, than what came before. Kevin Kelly saw this abundance coming, and tried to take seriously what marvels it would portend.

Except… well… when I ask my students “was Kevin Kelly right?”, they all get pained looks on their faces, glancing around the room to see who is going to step up and try to articulate the disconnect. What emerges is that Kelly was right about the abundance of content, but the abundance was beside the point.

You can see the tension emerge by focusing on the second and third paragraphs. Everyone isn’t busy uploading far more than they download (or, today, stream). Even the biggest bluesky sickos <raises hand> read far more than they post. A great many people still sit back and veg out, only occasionally contributing to the maelstrom. And the audience absolutely does matter online, because that’s where the money is.

What Kevin Kelly couldn’t see in 2005, because he was focusing on the new abundance, was that attention was about to become the dominant scarce resource. Hence, the attention economy. Hence Google/YouTube and Facebook becoming multi-trillion dollar businesses by algorithmically sorting attention and selling ads against the eyeballs.

He also couldn’t see that, once the money got big, content creation would stop becoming defined by scrappy amateur “prosumers,” and would instead become defined by the influencer industry. Hustlers would go on to make millions gaming the algorithms. Those are the defining features of the digital culture my students inhabit. The new abundance was just a starting point. The new scarcity was what defined the boundaries of both culture and industry.

This strikes me as a common mistake among tech futurists. It is relatively easy to tell a story that says (1) our society has been defined by this one scarce resource. (2) But technology is about to make that resource abundant. This changes everything. (3) Just imagine the wonderful possibilities once this new era of abundance has arrived.

They’re missing a critical fourth step: (4) With every new abundance, there will be a new scarcity. And how we manage that scarcity will tell us what sort of society we will be living in. Look toward the scarcity, not the abundance.

This all goes double for the AI future. SpaceX went public today.

Elon Musk is a trillionaire now, and his Space/AI/Social Media company is worth something like $2.2 trillion, with revenues of around negative $4 billion. Both Anthropic and OpenAI have filed to go public as well. Anthropic has had one profitable quarter, OpenAI has hahahahaha who are we kidding? OpenAI isn’t even trying to turn a profit. Why would they bother attempting to make more money than they spend? That isn’t how financial capitalism even works anymore.

These companies sell stories about the digital future to investors who give them money in exchange for shares of their storytelling company. These shares can be re-sold at a later date to some greater fool, often at a profit. SpaceX is worth $170/share today, not because investors broadly believe the company will generate several trillion in revenues, but because investors narrowly believe they will find someone who will buy their shares for at least $171/share at some later date.

The stories these companies tell about the digital future all revolve around abundance. And they are, if you have spent a few years studying this stuff, just the haziest, laziest stuff. Like, I’m sorry, “a country of geniuses in data center?” “We’re going to solve all of physics?” “Intelligence will be on tap, and then we’ll meter it like electricity?” The money has gotten so out of hand that these fellas aren’t even trying very hard anymore.

It would be nice if someone, anyone, could ask the AI barons this one simple question: If “intelligence” is about to become abundant, then what will the new scarcity be?” How will we organize ourselves in response to that scarcity? Who will profit from it, and who will be made more vulnerable as a result?

I can’t tell you what the world will look like in 2036. But I am damn certain that, with these guys in charge, it won’t be a utopia defined by limitless abundance. Of course it won’t. Let’s not be suckers, alright?

One final thought… zoom out far enough, and the history of the digital future looks something like this:

First, Silicon Valley built hardware that got smaller and faster and cheaper and better.

Then, Silicon Valley built software that became the mediating layer of the global economy.

And then, the finance industry ate the software industry. Every tech company today is a finance company.

Software ate the world, and then finance ate software, and now here we are.

It isn’t quite as simple as “after the Great Financial Crisis, all the finance guys moved to Silicon Valley so they could keep doing the same bullshit without all the regulatory scrutiny.” But that’s a close-enough approximation of recent history.

You cannot explain the actions of Silicon Valley today without paying attention to how the ludicrous excesses of financialization define the strategic behavior of the whole damn industry.

Terminal velocity

Dave Karpf on soon-to-be trillionaire (?) Elon Musk

Flaming Hydra

Hi folks

My latest piece for Flaming Hydra is a review of Ben Tarnoff and Quinn Slobodian’s new book, Muskism: A Guide for the Perplexed. It’s a very good book! You should read it! The book also provided a nice opportunity to think out loud about (1) what Musk’s success is doing to capitalism and (2) just how durable his empire seems to be. Here’s a taste of the review essay:

Please take a look at the review. (And please subscribe to Flaming Hydra! It’s the best!)

I’ll be back next week with another essay about the future, now, and then.

Best,

DK

David Wallace-Wells voiced some concerns about data center activism earlier this month, in a column titled “Data Centers Are a Distraction. The Real Fight Is Elsewhere.” Wallace-Wells writes:

It’s a fine question to ask. If you asked me to make a list of my concerns about the AI future, the excessive energy use from data centers would be on the list, but it wouldn’t rank among the top ten. Local fights against data center contracts are an imperfect vessel for the much broader AI backlash.

Loudoun County Data Center, photo by Hugh Kelly, PEC.

But there is also, I think, a compelling answer to his question. Data center activism is a bankshot. It can trigger a reset of the status quo, and has the potential to make a lasting impact on the digital future.

Let me provide a bit of scaffolding first: There are a few claims that I routinely make about digital futures’ past that really apply to the current AI bubble:¹

(1) The arc of any emerging technology bends toward money. We can have a better or worse digital future, depending on who sets (and who enforces) the rules. Giving the tech oligarchs free rein to pursue easy profit is a surefire way to end up in living in a dystopia. (extended argument here, from 2022)

(2) The essential thing to understand about tech futurism is that, when one of Sam Altman or Elon Musk’s predictions doesn’t pan out, they don’t have to give the money back. (extended argument here, from 2025)

(3) We’re in an AI bubble right now. How and when that bubble pops will dictate a whole hell of a lot about what happens next. (extended argument here, from 2025)

The single best thing that could happen to the AI industry right now is for the current finance bubble to pop. The sooner the bubble pops, the better. This is both because smaller finance bubbles are less calamitous than larger finance bubbles (replaying the 2000 dotcom crash is better than replaying the 2008 great financial crisis) and because, right now, Sam and Elon and Dario and Alex Karp and Marc Andreessen are calling all the shots. After the crash, their power will be constrained and contested. Decreasing their political power is just a fantastically pro-social outcome.

The trouble is that finance bubbles aren’t easy to pop. As Keynes supposedly once remarked, “the market can remain irrational longer than you can remain solvent.”

The standard approach is to loudly declare, with clear and compelling evidence, that the market has dramatically overestimated the value of a company (or a cluster of companies, in this case). That’s what triggered Enron’s downfall: a Wall Street Journal reporter raised the alarm, then Jim Chanos shorted the company and started making a public stink about their financial flimflam. The house of cards collapsed soon after.

That approach doesn’t seem to be working here. The Wall Street Journal keeps reporting on circular financing. Ed Zitron keeps shouting that the math doesn’t math. But media is so fragmented, the regulatory state so compromised, and wealth so concentrated that the investor-class just keeps pumping money in. The endgame seems to be that the AI industry achieves too-big-to-fail status, and that the major players can demand a bailout once the whole thing implodes. (That would be, y’know, really really bad. But it also seems awfully plausible, given the state of elite politics in 2026.)

Silicon Valley runs on futurity. The metaverse felt like the future for a couple years. So long as the metaverse had the aura of futurity, investor dollars flowed. Then the whole thing faltered and fell apart. The same thing happened to Web3 and a lot of other speculative technological futures. The AI industry knows this. It’s why the major players are constantly announcing new not-yet-half-baked products (quick, someone explain why the Jony Ive/Sam Altman partnership was newsworthy at the time) and constantly releasing manifestos. They need us to believe not just that AI is useful, but that it is inevitable, arriving fast, and bringing about a radically different world. That’s the only way to keep the cash-furnace stoked. The steady flow of investor cash only keeps going so long as they maintain a steady cadence of product demos, benchmark successes, and confident futurism.

Slow the pace down, and you lose the aura of futurity. And then the whole house of cards collapses.

Local opposition to data centers creates a ton of friction for the AI industry. They were not expecting this, and they don’t seem to know how to handle it. It’s a good leverage point for fighting the broader AI industry.

People really hate the AI future right now. And people have no place to put that hatred. They see it coming. They see Sam Altman and Elon Musk and Dario Amodei being smug assholes, promising that no one will have a job and only a few dozen people will have any money and teehee maybe it will end the world but probably not and anyway you better let them do whatever they want because otherwise China will do it first and where is your patriotism anyway? 

And they see that this handful of shitty dudes has all the money and all the power and AI is getting stuffed into every product and some of it is kind of cool, though a lot of it is annoying, but it isn’t as though anyone can do anything about it.

I heard this from my students all semester. They use AI, and they hate AI. This stuff is the future, and it’s a bad future, and they’re just kind of stuck with it.

Data Centers aren’t the perfect vehicle for all those sentiments. But they are a perfectly functional vehicle. David Sacks (or whoever replaces him as AI Czar) isn’t going to listen to you, because David Sacks (and that whole odious crowd) doesn’t care about democracy to begin with. And your member of Congress probably isn’t going to listen to you, because they’re worried about facing an avalanche of AI political spending in 2026. Also, even if they did, lol Congress doesn’t pass laws anymore, and any worthwhile law they did somehow pass would be vetoed by Trump or overturned by the Court.

But your state and local reps really might listen to you. AI isn’t primarily a zoning issue, but treating it as a zoning issue is one of the few ways to voice productive, collective dissent.

(It’s a bit like how copyright law became a multitool for stopping online harms. If someone posts deepfake porn of you on a social media platform, you are much more likely to get it taken down by saying “hey! copyright violation!” than by saying “are you fucking kidding me? I have rights. I am being harassed in the grossest possible ways. Your terms of service forbid this shit. Do something goddammit!” It isn’t a good sign that copyright is the only functional lever. But it is a lever nonetheless.)

People are driven to fight these data centers for a lot of reasons — energy bills, civic power, rage at the empty, anti-human imagined future that is being foisted upon us all. They’re all good reasons. This is how social movements get to work.

The original Luddites didn’t break machines because they liked breaking machines. They didn’t break machines because they were reflexively anti-technology either. (Read Brian Merchant, y’all) They broke machines because other forms of legitimate resistance were illegal or ineffective. Social movements gravitate toward whatever tools are available.

Local resistance to data centers is the best tool available right now. And the more I think about it, the more convinced I am that it’s a very good tool indeed. It has the potential to slow down the cadence of futurity, pop the finance bubble, and reset the entire public debate over how artificial intelligence is built, deployed, and regulated.

Data center activism is a bankshot. That’s what David Wallace-Wells doesn’t see. It’s what we all ought to recognize. There’s an opportunity here, and it’s a good one: The local fights create friction for industrial deployment. The friction slows down the cadence. Slowing down the cadence makes it impossible to sustain the bubble. And after the bubble pops, we gain much more agency to shape the trajectory of the digital future.

This is how we win.

Let’s take a look at the current state of the U.S. government:

Donald Trump, as President of the United States, effectively cannot be sued for anything at all.

Also, Donald Trump, as President of the United States, can sue his own government. For $10 billion. (Or eleventy-trillion. Really, for whatever number fits his mood.)

And he can then direct his Department of Justice to arrange a $1.776 billion settlement, so that the case never goes to trial. That settlement is just a slush fund, controlled by Trump and his lieutenants, providing payouts to his allies and supporters.

And, also, even if this is all blatantly unconstitutional, there are real questions about whether anyone would have standing to overturn it via the courts.

And that’s because, in an at-least-barely functional democracy, the appropriate response is for Congress to initiate impeachment hearings. On paper, that’s the remedy: Congress impeaches, and/or Congress passes legislation forbidding this outright graft and corruption.

But, ell oh ell, this particular Congressional majority would never.

On paper, the remedy for a Congressional majority that would never rein in audacious graft and corruption is to vote the fuckers out of office six months from now.

But, ell oh ell, this particular Supreme Court majority has decided to lock in Republican gerrymanders and forbid/delay Democratic gerrymanders. No matter how unpopular the Republican Party is in November, the election is likely to be very close. And that’s assuming we have free and fair elections. (Which… yeah, let’s hope for the best but plan for the worst, y’know?)

Oh and another thing: to remove Trump from office over the outright graft and corruption requires both a majority in the House and a 67-vote supermajority in the Senate. It is almost a mathematical certainty that Democrats will not have a 67-33 majority after the 2026 election. They would have to win almost every Senate seat, even in the reddest of states.

Which means, so long as Trump maintains control of his own party elites (in Congress and in the courts), in practice there is nothin to stop even the most brazen corruption. Because both the Republican Supreme Court and the Republican Congress have decided to render him immune from the law, and have also rendered each other immune from the electorate.

And, for a corrupt politician who can only be deposed by losing control of his own party elites, it is pretty useful to have a couple billion dollars in a slush fund to reward your allies (there’s the carrot), while simultaneously directing the FBI and DOJ to harass your enemies (there’s the stick).

None of this ends well. That much is obvious. But it also doesn’t end through the normal checking-and-balancing of the U.S. government as it was designed to work on paper. This, too, ought to be obvious at this point.

The frustrating thing is that it also will not end soon. There is no straight line that I can draw from the outrages of this week to the reconstruction that must certainly, one day, follow. We are well and truly off the map. Here be dragons.

And that temporal delay is, among other things, completely exhausting. The appropriate emotional reactions to learning that Donald Trump’s government has settled a Donald Trump lawsuit by giving Donald Trump $1.776 billion to hand out to his friends and supporters are (1) endless screaming into the void until you go hoarse, (2) incoherent sputtering, followed by your choice of self-soothing behavior, and (3) throwing your hands up in disgust and deciding to tune out politics entirely.

Option 3 is, of course, the path to helpless acquiescence. Giving up and tuning out is what they want. And it is much better not to give it to them. Don’t do that.

But also, let’s be honest: we can see the appeal, right? Because this is so corrupt and so absurd, and someone in power surely ought to do something, but the arrangement is self-reinforcing in a way that makes it unlikely to survive in the long-term, but nearly impossible to dislodge in the short-term. And so we are left to ponder when, exactly, the long-term is supposed to arrive, and why, precisely, the people in power aren’t doing more.

The news next week will likely be even more corrupt and more absurd. Empirically, we keep finding out that we haven’t hit rock bottom yet, because if we hit rock bottom then there would be some sort of noticeable recoil on impact.

We will, eventually, rebuild. But it will not be a return to normalcy. Once an institution has been this compromised, once it has been brought down through the uneven application of its own rules, it can no longer be merely restored. It has to be amended and restructured.

In the meantime, we are left to scream into the void. We also use this time to organize, and to prepare for the future. There are more-productive and less-productive ways to live through a collapse. We ought to use this time to as best we can.

But the endless screaming is part of it, too.

It’s the appropriate emotional response.

Eighteen months ago, the metaphor I was using for Bluesky was that it was the Ship of Theseus. This was during the post-election haze, when people were abandoning Twitter X in droves, and turning to Bluesky. Piece by piece, all the parts of old-Twitter that people actually liked were being reassembled on Bluesky.

I still think there’s something to that. It is conceptually pretty fascinating that Bluesky works at all. As I mentioned in the kicker of that piece, “Seventeen years of network effects should, in theory, have been an insurmountable advantage. But network effects and habitual user behavior have proven no match for the strongest repulsive forces in the known universe: Elon Musk’s personality.”

But it’s also true that the Bluesky migration was incomplete. Last month, in a post titled “Is Bluesky dying,” James Ball sounded an alarm, pointing out that daily activity has fallen substantially over the past year. Earlier this week, David Roberts made a similar point, noting the net-negative effect of politicians and media institutions remaining on X and being dragged further right.

From James Ball, “Is Bluesky dying?”

I don’t think Bluesky is dying, per se. But I do think we’ve seen in the past year that Bluesky is quite niche. My initial response to Ball’s argument was “The entire web is dying. Users aren’t going from BlueSky to another site (x/insta/threada/tiktok). Users are going to chatbots.”

Okay so here’s my actual Bluesky-is-dying hypothesis:

The entire web is dying. Users aren’t going from BlueSky to another site (x/insta/threada/tiktok). Users are going to chatbots.

I know traffic to news sites has cratered (like 90%). My hunch is traffic to all the social platforms is down too.

— Dave Karpf (@davekarpf.bsky.social)
1:33 PM • Apr 4, 2026

Petter Törnberg has since published a paper in JQD that fills in the picture with relevant data. Social media engagement is down everywhere. We don’t know for sure that they’ve turned to chatbots. But it’s pretty clear that people aren’t leaving the progressive text-based social media network for an alternative text-based social media network.

Is social media dying? How much did Twitter change as it became X? Which party now dominates the conversation?

Using nationally representative ANES data from 2020 & 2024, I map how the U.S. social media landscape has changed

Here are the key take-aways 🧵

Full paper out now in in JQD:DM!

— Petter Törnberg (@pettertornberg.com)
6:13 AM • Apr 22, 2026

In a nutshell, the ANES data shows:
📉 Social media use is shrinking; engagement collapsing
💥 Twitter/X posting has moved ~70 POINTS to the right
🧩 Platforms are splintering
🔊 Fewer people are talking — but those still talking are more politically extreme

— Petter Törnberg (@pettertornberg.com)
6:13 AM • Apr 22, 2026

That doesn’t mean Bluesky is going to die. But it does mean that the site isn’t ever going to replace peak-Twitter. Nothing will.

II think it’s helpful to introduce another metaphor. The more I’ve thought about it, the more I’m convinced that Bluesky is a record store.

John Cusack, in arguably the perfect John Cusack film.

Record stores were once an essential part of the music industry’s attention economy. In the 1990s, you could access music through (1) the radio, (2) MTV, and (3) the record store. The record store was arguably the most important of the three, because it was the only point-of-sale location. Record stores were pretty big businesses (remember Tower Records?) and cultural institutions (Empire Records, High Fidelity).

Then came iTunes, Spotify, and Amazon.

Record stores still exist in 2026. Some of them even thrive. People like me like record stores. They remain active, niche cultural institutions. But they are distinctly different entities than they were in their heyday. They have a smaller footprint, and they operate within those limitations.

The same is true for text-based social media platforms. The era of peak social media is over. Twitter and Facebook used to be the “attention backbone” of the entire internet. (Ben Smith’s book, Traffic, is probably the best discussion of how media worked in this time period.)

People like me also like Bluesky. In fact, I’d go so far as to say that people who like record stores all pretty much like Bluesky. And that’s probably the right frame of mind for thinking about text-based social media in general. It isn’t a dominant media institution anymore. It’s niche. It has some cultural cachet, but you will be disappointed if you expect it to recapture the glory days.

The mid-2020s is an awkward time to have launched a text-based social media network, in much the same way that it is an awkward time to open up a new record store. This isn’t exactly the future anymore, y’know?

And/but, I think one reason why people keep having existential debates about the present and future of a niche social media network is that there are two types of deep uncertainty about what the future will look like.

(1) There’s AI-uncertainty: social media doesn’t drive web traffic anymore. Nothing seems to have replaced it. That’s, y’know, bad. There’s a giant hole at the center of the attention economy, and it’s getting bigger, and everyone is vaguely nervous about it.

(2) Then there’s authoritarian uncertainty: As David Roberts pointed out, all of our mainstream institutions are sprinting to the right. Everything is bad and depressing. Everything is falling apart. It sure would be nice if the spaces where we congregate online felt as vibrant and productive as they used to.

All that uncertainty is real, and worth dwelling upon. But I think it’s helpful to calibrate our expectations. Bluesky isn’t dying, but it is attempting to thrive within a declining niche. Bluesky isn’t the reason why mainstream institutions are galloping to the right, but it is less useful as a space for effective counter-mobilization than it would have been, back when text-based social media was the backbone of the entire attention economy.

I’m going to happily keep posting my little jokes and rants and book reviews on Bluesky. It’s a nice distraction, and purpose-built for people-like-me. But most of my students will probably never get into it… and the ones who do will probably be the ones who hang out at record stores.

Meanwhile, the hole at the center of the attention economy keeps expanding, and the fascist overlords get ever more brazen, and everyone tries to figure out what we can do about those twin crises.

I’m not sure how we solve them. But I don’t believe these are record store-shaped problems.

Back in 2022, when Elon Musk started making moves to acquire Twitter, he made a paperwork error. The Securities and Exchange Commission (SEC) requires large investors to file a disclosure statement within ten days of acquiring 5% of a publicly-traded company’s stock. Elon waited twenty-one days to file his disclosure. This let him buy a lot more shares in secret. As soon as his stock-buying spree was made public, the price of the stock shot up. Elon saved around $150 million by delaying this disclosure.

Of all Elon’s crimes, this is among the least serious. It’s also maybe the simplest. Everyone agrees you have to file the form. Elon didn’t file the form. Elon saved $150 million from this little “oopsie.” So the SEC assesses a penalty of at least a bit more than $150 million, Elon pays the fine, nothing to see here, please move along.

But of course this is the 2020s, and this is Elon we’re talking about, so…

Elon jumping at a rally. Image from NYT article

The SEC investigation took nearly four years. FOUR. YEARS. Some of that was probably the SEC’s fault. They could have moved faster. But much of it was Elon’s lawyers doing elaborate lawyer-things. If you pay your lawyers by the year, you can avail yourself of a whole host of procedural opportunities to drag things out.

And that meant the case was handed off from Biden’s SEC to Trump’s SEC. So (yadda yadda yadda) Elon’s lawyers have negotiated a settlement. Elon admits no wrongdoing, but agrees to pay a $1.5 million penalty via a revocable trust. That’s effectively a 1% tax on the money he saved through his disclosure failure.

As Matt Levine put it:

This creates an obvious perverse incentive. The SEC rules don’t meaningfully constrain anyone’s behavior if there’s no real penalty for violating them. And also keep in mind that this is a perverse incentive that only applies to the super-wealthy. I am not in a position to buy 5% of a publicly-traded company, so am in no position to benefit from the collapse of this particular SEC rule.

But, even worse, this paperwork-error-loophole doesn’t even apply equally to all of the super-wealthy. Because in 2026, the Administrative State is now solely in the business of doing favors for Donald Trump’s allies and creating problems for Donald Trump’s enemies. Elon can break the disclosure rules, scoop out the profit, and then pay a 1% courtesy fee. Habitual Democratic Party donors ought not try this at home.

And that’s where the effects of Citizens United come into play. The Supreme Court decided that everyone should be able to spend as much as they like to influence U.S. elections. (Election experts warned “uh, hey, this is going to lead to some truly outrageous corruption,” and the Court said “what? LOL I’m sorry I can’t hear you over the sound of this private flight I’m taking with Harlan Crow.”)

The direct impact of Citizens United is that, if you are a super-wealthy individual facing multiple government investigations, the cheapest thing you can do is try to buy the government outright.

Elon Musk’s paperwork error was the simplest and smallest of the government investigations into his business empire. He spent $290 million on the 2024 election. That’s an outrageous, unprecedented sum. And it’s also, for him, a bargain. Had Kamala Harris won, he would have remained on the hook for billions in fines. It was cheaper for him to buy the government outright than to pay what he owed.

This feels like a simple parable of life in the United States in the 2020s. A handful of billionaires own everything. They are sloppy. They don’t bother to obey the rule of law. They profit as a result. And there is nothing — not even the government — that they cannot buy.

It’s a story of simple corruption. It cannot last. But it is not self-correcting. This sort of everyday corruption keeps spreading until everything collapses. Systems like this break down, and then have to be rebuilt through tremendous collective effort.

It’s all just so… pathetic. This was Elon’s smallest infraction, and even in this case, he refused behave as though the social contract applies to him by just paying the damn penalty. And he got away with it, because of course he did, because that is how the Supreme Court and the Trump Administration have chosen to arrange the incentives.

Photo by Braedon McLeod on Unsplash

On the most recent episode of Hard Fork, Casey Newton and Kevin Roose were discussing the data center boom. They had some good points, but also made what I believe to be a crucial error:

Newton elaborates on this point in a post last week, “We may now know what kind of AI bubble this is (Think railroads, not crypto).” The railroad analogy isn’t a bad one, but I think it’s worth dwelling upon for a bit. I have three notes (all of which are sort of a remix of a 2025 piece titled “It’s Giving Enron.”):

(1) The AI bubble is, indeed, not like the crypto bubble. I agree with Casey on this point. The fundamental value of bitcoin is basically zero. There are interesting reasons why bitcoin hasn’t fallen to zero (and likely will not anytime soon), but all of those reasons are variations on the theme of “lolwtf the entire economy is just gambling and speculation now.” Bitcoin isn’t valuable because it has uses that generate economic value. It’s valuable because investors have agreed to attach abstract value to it as a speculative asset.

AI has several uses that generate economic value. The industry as a whole is a cash furnace, but while the too-simple explanation of crypto is “what if tulip bulbs formed the basis of a multitrillion-dollar speculative economy?”, the too-simple explanation of AI is “what if we introduced machine learning techniques into everything everywhere all the time?” Both are bubbles, but Casey is correct that they’re materially different sorts of bubble.

(2) Railroads are a decent analogy. But broadband and the 2008 housing crisis are closer to home.

Newton writes:

He has me in the first half. Yes, we’re looking at massive overbuilding and speculation, likely followed by a crash and flameouts. Yes, real underlying products are likely to emerge and flourish after the crash. The analogy I’ve leaned on in the past is the Web 1.0 broadband buildout-and-crash (It’s giving Gilder). I still suspect we’re that some of the exceptionally fancy accounting that we’re seeing today will, after the crash, be cast as at least borderline fraudulent (hence, it’s also giving Enron).

But the outstanding question is just how economically transformative the underlying product turns out to be. Railroads pretty completely transformed the economy. Broadband is no flash in the pan, but it is smaller by comparison. The underlying question isn’t is there value in there somewhere (yes, there is), but how much value is in there, relative to the debt levels? 

And that’s what brings the 2008 housing crisis to mind. The great financial crisis nearly cratered the entire global economy. That wasn’t because the underlying asset class had no intrinsic value. The underlying asset class was houses! People live in those! We still don’t have enough of them today! But the problem was that they financialized the hell out of the housing market, and created absurd subsidies for unreliable home loans, just so they could spin up and sell shoddy financial products. If you build a financialized Rube Goldberg machine on top of a limited-but-valuable product, things can go very, very wrong.

(3) And we are still in the free-trial-period of most AI products. Everything is being subsidized. That renders a false picture of the demand curve.

The bit from Hard Fork that really set me off was when Kevin Roose dismissed critics as saying “these AI companies would never be able to generate the demand to pay for all of the expensive data centers and infrastructure projects they wanted to do. And now [the critics have] shifted to, well, there's so much demand. what if they can't build enough to support the demand they have?”  

Roose is conflating two different types of demand here. He ought to know better.

Here are two statements that can simultaneously be true: (a) the AI industry faces extraordinary and growing demand for compute. It does not have enough data centers to satisfy this demand. (b) AI users pay for practically none of that compute, and would stop using AI if they had to pay the direct costs.

Earlier this semester, one of my advisees told me about one of his Claude Code projects. He asked Claude to build him a study guide with flash cards to help him prepare for a biology exam. I am generally an edtech skeptic, but this seems fine to me. Creative, even. He isn’t having Claude do the reading or the writing or the thinking for him. He’s just vibe-coding a test prep assistant to help with rote memorization tasks. Good for him.

How computationally intensive was his little vibe-coding project? I don’t know. Neither does he. He doesn’t need to.

The current deal (circa two months ago) is that my student pays $20/month to Anthropic, and Anthropic lets him build little things like this. My student doesn’t have to worry about how computationally intensive the exercise is. The cost of computation is borne by Anthropic, not the student.

If it had cost my student $500 to build that study guide, he would have just written his own damn flash cards. If it cost him an extra $20 on top of the monthly subscription, he might have gone the index card route as well. When the demand for compute is subsidized by the same AI industry that is trying to raise fresh rounds of investment (based on a pitch deck that highlights skyrocketing computational demand), we end up with a heavily distorted picture of the demand curve.

Could this vibe-coding project be more computationally efficient? Almost certainly. This is the ongoing threat that Deepseek poses to the AI industry. Deepseek’s model can do basically everything Claude does, at 1/6th the cost. With the exception of Deepseek, the AI industry has taken a scale-at-all-costs approach.

One of the major reasons why existing data centers cannot keep up with the computational demand is that the AI industry is not the least bit interested in pursuing efficiencies that would make their products less computationally demanding. They’ve constructed a sort of financial flywheel, where they subsidize computation-heavy behavior from users, then report back to investors about the soaring product demand and use it to justify the next investment round. If they built more efficient AI models, then it would paradoxically become harder to attract the next trillion dollars of investor cash.

This industry-wide subsidy system can’t go on forever. The money, eventually, has to come from somewhere. The Verge’s Hayden Field reported a couple weeks ago on the “AI monetization cliff” that seems to be approaching. “After years of offering cheap or totally free access to advanced AI systems, the bill is starting to come due — and downstream, users are beginning to feel the pinch.” Field is right about the current trends. But we shouldn’t be surprised if the AI companies balk and decide to hold off on profitability for just a little longer (aka indefinitely).

You’re about to feel the AI money squeeze

The AI monetization cliff approaches.

The Verge • Hayden Field

Back in 2022, I wrote a piece for The Atlantic titled “Money Will Kill ChatGPT’s Magic.” I think that piece still holds up quite well. The trajectory of the AI industry is going to bend toward money, and so long as we are in the free-(or cheap)-trial-period, we will have a heavily distorted view of what sort of AI products are actually profitable. All the major AI companies are reporting massive growth in their subscription plans, while also facing the obvious problem that those subscriptions are revenue-negative. It has been three and a half years, and the industry is still dramatically subsidizing its products and then bragging about all the demand.

Money Will Kill ChatGPT’s Magic

Buzzy products like ChatGPT and DALL-E 2 will have to turn a profit eventually.

The Atlantic • David Karpf

All of which is to say that AI companies have not generated the demand to pay for the expensive data centers. They’ve generated and subsidized the demand for compute, and then used that computational demand to make a case for further AI infrastructure investment.

The crash occurs when we find out how much consumers value AI when they have to pay the full sticker price for the unsubsidized products. If it ends up that we merely have oversupply for the worlds-most-valuable product, then a few companies go bankrupt while the global economy roars forward. But if it turns out that the financial flywheel massively overstated how much economic value we get from AI, then all those trillions the banks and governments handed over to Sam and Elon might capsize absolutely everything.

I think it’s a mistake to pretend that AI is completely useless. But it’s an even bigger mistake to treat the subsidized compute demand as proof that a once-in-a-century economic transformation is currently underway.

[First, a programming note: I recorded a podcast last week with Flaming Hydra Maria Bustillos and Joe Macloud last week. We made fun of Alex Karp and Palantir’s goosestepping manifesto. We covered a lot of ground in this one, of the sort that all “Future Now and Then” readers are likely to enjoy. Check it out]

There’s an essay that I started drafting in the before-times (2019), and never got around to finishing. It has been rattling around my head this week, as I watch the Supreme Court once again defenestrate its own institutional legitimacy.

Trump’s first Director of National Intelligence, Joseph Maguire, was testifying before a House committee. There were a lot of questions about foreign election interference, and whether our voting machines were secure from hacking. Some of these questions came from people online who didn’t know anything. But some of them also came from people like Bruce Schneier, who know a lot about digital security and had some very real concerns about vulnerabilities in the system.

Maguire’s response focused on public faith in the validity of our election results. Basically, he said that the intelligence community’s #1 job was to make sure Americans had faith in the election results. I remember hearing that and thinking “no, that’s wrong. What are you doing, man?…”

The intelligence community’s #1 job (circa 2019. Simpler times..) wasn’t to maintain American faith in election results. That’s a communications campaign. You hire people like me and the graduates of my program for stuff like that. The intelligence community’s #1 job should be making sure the election results aren’t interfered with!

I drew up a nifty little four-box chart to illustrate:

Four-box charts are my passion.

From a comms perspective, let me assure you that it is much easier to increase public trust in an institution when that institution is behaving in trustworthy ways. It still takes work — particularly when you’re facing a media system that rewards lies and misinformation — but the work is an order-of-magnitude less challenging when the facts are all on your side.

It’s also a different type of problem. If your elections are compromised, but people don’t believe they are compromised, then you will maintain trust in the institution (at least for awhile. It will not last.). If people believe your elections are compromised, but they aren’t, then you will face social instability (cough, cough, January 6th). It worried younger-me that our Director of National Intelligence seemed more worried on the optics of legitimate elections than on the work of safeguarding electoral legitimacy. The people in positions of authority were absolutely focused on the wrong things.

I bring this all up in light of the Supreme Court invalidating what was left of the Voting Rights Act earlier this week.

I don’t have to tell you this is bad. We all know it’s bad. Louisiana’s Governor has announced he will suspend the state’s primary elections (even though voting has already begun) so he can redraw the district boundaries to benefit Republicans and disenfranchise black people. For the foreseeable future, it’s going to be all-gerrymandered-districts-all-the-time. Elected officials get to pick their voters, not the other way around. (Don Moynihan’s essay on this topic is a must-read)

I also don’t have to tell you it’s unsurprising. The six-member Supreme Court majority is comprised of unabashed partisan operatives, especially when it comes to elections. The Supreme Court’s institutional legitimacy is like a phantom limb. It vanished years ago, but we still carry on as if it was still present. The Roberts Court did what the Roberts Court does. The Court is a partisan institution now.

As Jonathan Ladd notes, this has unavoidable implications for democracy reform proponents. This Supreme Court majority will ignore the plain text of the Constitution in order to achieve its political aims. You can’t fix our political system through legislation if Roberts, Alito, Thomas, Kavanaugh, Gorsuch, and Barrett will just strike down any legislation that disagrees with them.

And yet still there are people-who-should-know-better, insisting that we ought not undermine public faith in the Supreme Court by stating plainly and repeatedly what this Court is doing.

This, I suspect, is for two reasons. One is that there simply aren’t many Supreme Court management strategies that don’t require audacious and unlikely reform efforts. We aren’t going to norms our way or shame our way into Alito and Kavanaugh becoming responsible jurists. They have told us, repeatedly, who they are. They have lifetime tenure, and an axe to grind. It would probably require court-packing or a constitutional amendment, and that’s a tall order with a gerrymandered House and a geriatric Senate stacked with head-in-the-sand traditionalists.

The other reason is social stability. We want the public to have trust in institutions. If the public loses faith in the court system, that’s a precursor to many bad things happening. Here, I drew it up as another four-box chart:

I said four-box charts are my passion. I didn’t say they were my skill.

If it was still 2019, then there would perhaps be room for debate about the objective facts of the judiciary. Citizens United and Shelby County were real bad. But the overall record of Supreme Court behavior left room for some plausible deniability. They were still pretending back then, and they had not yet secured their six-Justice majority.

But what stands out to me is that the problem with the Court is not a comms problem. People have lost faith in the Court because they have paid at least a modicum of attention to what the Court does. This is not a problem that can be messaged or massaged away. And insisting at every turn that we must not undermine public trust in institutions has only contributed to the people in power’s sense that corruption has no consequences.

American Democracy will not be saved by a landslide election in 2026. A landslide election in 2026 (if it happens) will only be a precursor to the incredibly hard, necessary work of structural democracy reform. It will, at best, be a window of opportunity.

The same is true of 2028. It would be so nice if all that we needed to recover from Trumpism was to win two consecutive elections and then get back to normal. But that’s a fantasy. We will not fantasize ourselves out of this state of affairs.

So I find myself, on weeks like this one, thinking about phantom limbs and the work it takes to adjust to new realities. I find myself thinking about multi-member districts in the House, radically restructuring the Senate, and turning the Supreme Court into a rotating panel of judges — all of whom have term limits and ethics laws with real teeth. I think about wealth taxes and rewriting all the laws that have given the super-wealthy immunity from the rules that govern the rest of society.

These are all impossible dreams in the system that we have today. I cannot draw a straight line from where we are today to where we would need to be. But the system we have today is awash in corruption, and absolutely everyone knows it. None of this can be solved through optimistic messaging and can-do spirit. We might as well start thinking about what we replace it with, because the center, as-is, cannot hold much longer.

I kind of think every time a tech CEO publishes an online manifesto, they are a little bit doing it just to annoy me. That goes double for Palantir CEO Alex Karp. (The “F” in my last name is silent. Stop beclowning our shared last name, Alex!)

I have a new piece up at Tech Policy Press, discussing and dissecting Karp/Palantir’s latest manifesto. A couple key passages:

A few snarky comments that I left out of the piece, since I like to class it up when I write for serious outlets:

-Alex Karp looks like Taika Waititi after a three-day bender. That’s mean and petty of me to say. But if a billionaire is gonna go full mask-off fascist, then I’m gonna be a little mean and petty.

Alex Karp at the Dealbook summit. Having a normal one.

-Palantir is a real defense tech company, with significant government contracts. But it also has a price/earnings ratio of 230, which absolutely makes it a meme stock. For people wondering why Palantir would publish a garbled, authoritarian manifesto like this/right now, that seems like the answer. Expect the company to do a whole lot of “look at meeeee” shit to justify their outrageous valuation. None of that shit will age well, but memestock companies don’t worry about the medium term. They just string together absurd promises to make it through the next quarter.

-Like the Andreessen Manifesto, part of what Karp and Palantir are doing here is telling potentially employees “this is the party line that you have to believe if you want to work here.” It’s garbage writing, and garbage thinking. But there’s also power in it. This is all a downstream consequence of giving some of the worst people in the world access to practically unlimited capital.

Anyway, the point is that Alex Karp is the worst, and we should cyberbully him into changing his last name. Here’s a link to the Tech Policy Press piece, which has more substantive analysis and insights:

Palantir's Manifesto Is as Subtle as a MAGA Hat

Palantir's 22-point manifesto is nonsense, writes Dave Karpf. But it at least it is short enough to be clarifying, he says.

Tech Policy Press • Dave Karpf

Someone threw a molotov cocktail at Sam Altman’s house this weekend. The house is fine. Altman is fine. Altman’s family is fine. The suspect appears to be an Existential Risk guy — someone who read Eliezer Yudkowsky’s book and took it too seriously.

Political violence is bad and counterproductive. One of the reliable, immediate impacts of political violence is a rush to vocally discredit whatever movement the political violence is associated with. So it matters, in moments like this, what conceptual boundaries we place around this phenomenon.

In a substack post yesterday, tech writer Jasmine Sun cast it in terms that I find deeply frustrating. The post is titled “AI Populism’s Warning Shot.” Here’s the key passage:

According to this definition, the 20-year-old throwing a molotov cocktail at a mansion gets lumped together with the community activists attending zoning board meetings to protest AI data center construction in their neighborhood. According to this definition, both I and Eliezer Yudkowsky (as people focused on AI’s unwelcome societal changes, rather than ChatGPT’s personal utility) are members of the same movement.

That’s ridiculous. Resist the urge to lump all forms of AI criticism into the same umbrella category. It obscures much more than it reveals.

I would sorely like to nip this in the bud before it becomes a thing.

I saw a similar mistake in 2016-2017, as the research community struggled to make sense of Trump and Brexit. Was Donald Trump better understood as a “populist” or as an authoritarian. He fits both definitions. And the built-in advantage of calling Trump a populist was that it provided cover. There are left-populists, like Bernie Sanders, and there are right-populists, like Donald Trump. Populism is a distinct rhetorical tradition. Researchers can talk about the dangers and benefits of “populism” without worrying that they will sound too partisan or outlandish. The disadvantage of treating Trump as a populist was that it obscured much more than it revealed. Trump and Bernie Sanders aren’t two sides of the same coin. There is safety and comfort in hiding what you mean, but there isn’t much clarity in it. It took years of internal scholarly debate before the field at large got comfortable just saying outright what we meant.

What a pointless waste of time that all was.

Here’s another relevant passage from

This is, I think, correct on the merits. AI is now a significant chunk of the economy. The industry has succeeded in stuffing AI into every product and every consumer experience. And that means AI policy is no longer the remit of a small, expert policy community. That was an early times phenomenon. The early times are over. This is what success for the industry looks like.

All the more reason to avoid the instinct to lump all forms of AI resistance into the same category. The Yudkowksy fans throwing molotov cocktails are a distinct discursive community, operating far outside the boundaries of what we might call “normal politics.” The data center challenges are mostly coming from what I would label Sierra Club-types. These are people showing up to community board meetings, writing letters to their legislators, making demands about industrial transparency and reporting, pollution and energy prices.

These two variants have practically nothing in common. The Yudkowsky crowd literally staged a protest this February in favor of billionaires. The Sierra Club types are deeply distrustful of corporate power, and deeply committed to participatory democracy. Bundling them together under the heading of “AI populism” makes it even harder to reach analytic clarity. The sole thread that ties them together is that they both are a problem for the AI industry.

Yudkowsky’s friend group. They are all so weird. Photo by Abigail Van Neely, of MissionLocal

Altman responded to the molotov cocktail incident with a heartfelt plea to turn the temperature down. He blamed last week’s Ronan Farrow/Andrew Marantz article, “Sam Altman May Control Our Future — Can He Be Trusted.” This is a textbook strategic communication play: the molotov cocktail-thrower discredits the entire movement he is associated with, so Altman is associating him with absolutely everyone who voices concerns.

Tech writers shouldn’t be in the business of helping him paint with such a broad brush. There’s really no need to do so.

Ohandbytheway the reason this quite matters, the reason why companies like OpenAI ought to be aggressively in favor of procedural democracy, is that there isn’t a “no backlash of any sort” option available. This isn’t solely the product of too much artificial general intelligence alarmism. This is, for the most part, the necessary byproduct of the size and scale of the industry. You don’t get to bite off this much of the global economy without facing serious questions about what comes out the other end (h/t Henry Farrell).

The options for the AI industry are (a) legitimate backlash or (b) illegitimate backlash. The mass public turns to violence when the avenues for legitimate, orderly contention are unavailable or non-functional. The Luddites smashed machines because it was illegal to forming unions.

I suspect the next few years will see an awful lot of anti-data center activism. People are going to raise their voices and say “we don’t want these data centers raising energy prices here. No more giveaways to Musk/Altman/Zuckerberg.” OpenAI’s lobbyists and comms consultants will surely brand it the “new NIMBYism,” and “AI populism.” They’ll treat participatory democracy as a form of damage and try to rout around it. I expect they’ll be quite cutthroat in their maneuvers.

That’s a mistake though. The friction of participatory democracy creates a pathway for legitimate resistance. If you do away with that friction, the illegitimate alternative you’re left with is firebombs. As I’ve written elsewhere, Democracy is an incredibly good deal for elites, one that they ought to stop taking for granted.

There is a broad sense right now that tech billionaires run the world, entirely unconstrained by the public. This, to a great degree, is because they do. They bought the government, shredded the regulatory constraints, and treated neo-feudalist edgelords as political sages. It was a short-sighted maneuver, destined to fail. Silicon Valley ought to be more appreciative of the social stability provided by democracy. The alternatives are so much worse.

This week I read Pepper Culpepper and Taeku Lee’s new book, Billionaire Backlash: The Age of Corporate Scandal and How It Could Save Democracy. It’s quite a good book. I will be assigning a couple of chapters to my class next fall.

The central thesis of the book is that scandals — particularly corporate scandals that do not fit easily within pre-existing partisan boundaries — can be drivers of significant political change. Scandals can activate latent opinion, moving political leaders to action. Culpepper and Lee are gifted political scientists, and they provide strong historical and cross-national evidence in support of the argument.

It is ultimately an optimistic book. Culpepper and Lee see reason for hope that scandal and corruption will generate the conditions for democratic renewal (what they term “good populism”). We are living through the years of DOGE and RFK Jr’s assault on public health. The Trump administration is inventing new kinds of insider trading, and refusing to prosecute audacious white collar crimes. This stuff is not popular. It’s bizarre. And it has real consequences. These are the sort of conditions that can produce large-scale change, particularly when it breaks partisan containment.

The book strikes me as an extension and update on the framework that John Kingdon established with his 1984 classic, Agendas, Alternatives, and Public Policies. Like Kingdon, Culpepper and Lee are effectively asking “when does an idea’s time come?” And like Kingdon, they arrive at the importance of attention-shaping moments, along with the critical role of policy entrepreneurs. A lot has changed about politics in the intervening decades, so it’s nice to see an update of this sort.

That being said, I have a couple of critical notes to share.

I think a fair two-sentence summary of their argument would be something like “when a scandal gets picked up by 60 Minutes, politicians feel pressured to respond. And if that scandal isn’t already partisan-coded, that response may significantly alter the status quo.” Hence the title, Billionaire Backlash — when there is a major scandal with billionaires and massive corporations as the villain, we have ripe conditions for significant backlash.

It is a book, in other words, focused on public backlash to the billionaires.

I would argue that there is another, newer phenomenon currently warping our political system though: a backlash by the billionaires, against the public. The role of money in politics is not new, but (as I noted last month) the scale of billionaire influence operations is unprecedented.

Billionaire Backlash is the culmination of a decade-plus of research by Culpepper and Lee. This is both a strength and a weakness. They collect original data, across multiple cases and multiple countries, all of which provides robust support for the overall argument. But the focus on systemic patterns downplays the ways that our current moment really is different.

Just in the past year, Larry Ellison bought CBS for his son David, and they put Bari Weiss in charge of the whole thing. The Ellisons also control the TikTok algorithm, and are taking over CNN and HBO as well. Elon Musk bought Twitter, Mark Zuckerberg runs Facebook, Jeff Bezos now basically pre-approves all Washington Post editorial stances, etc. This level of hands-on billionaire media control is just about unprecedented in U.S. history.¹

It seems to me that this matters a lot for Culpepper and Lee’s model. Scandals only happen when the mass public is exposed to new, non-partisan-coded information through programs like 60 Minutes. It is historically the task of policy entrepreneurs to shepherd these controversies at least until they arrive at that point. What happens when the billionaires outright purchase 60 Minutes and instruct Bari to revamp the program?

To their credit, Culpepper and Lee devote a few pages of the final chapter to the problem of billionaire media ownership and social media fragmentation. They have seen what Musk did to Twitter, and what Bezos did to the Washington Post, and they warn that this is absolutely a problem that needs solving. But, speaking as a media scholar, I found their proposed solutions to be a little halfhearted. They suggest using the threat of government intervention to “compel a more truth-based public sphere.” Something like NewsGuard, perhaps.

In 2024, I reviewed The Death of Truth, by Newsguard’s Steve Brill, for the Journal of Democracy. NewsGuard is a fine endeavor. It is not nearly up to the task of counterbalancing the billionaire assault on the press. It’s not even close.

This is a hard problem. We shouldn’t don’t expect Culpepper and Lee to solve them for us. But it might be the hard problem that proves ultimately fatal for democracy. So it kind of puts a damper for me on their ultimate optimism.

In a still-functioning democratic system, backlash to the billionaires is a source of civic renewal. In our currently-existing political system, backlash by the billionaires threatens an end-run around that process. The book is quite worth reading, but it allays few of my pessimistic concerns. Democracy requires a free and (at least semi-)independent press. The plutocrats went ahead and bought the media system. Whether they will be able to preempt backlash to the billionaires remains to be seen.

Today I want to revisit an old prediction. I got this one wrong. And I think the relevant lesson for 2026 is that I am still not nearly jaded enough.

One of the definitive asymmetries in our modern legal system is that some people pay their lawyers by the year, while the rest of us pay our lawyers by the hour. A good rule of thumb is never underestimate how much the people who pay their lawyers by the year can get away with.

In March 2023, I wrote a post on Substack titled “How Long Does Twitter have left?” I gave it about six months. That… has not happened. Ex-Twitter isn’t going great, but it is still extremely here.

My reasoning at the time was that the publicly-reported finances were atrocious, and the underreported finances were even worse. Twitter was sitting on two financial time bombs — the pending employment lawsuits, and the inevitable regulatory fines. Once those hit, it seemed likely that Elon would make a face-saving exit.

Here are the relevant passages from three years ago:

I wasn’t wrong about either of those financial time bombs. But I was, in retrospect, pretty naive about the time horizon.

The lawsuit from the former Twitter executives finally settled in October 2025. The lawsuit by former Twitter employees who Elon decided to just… not pay their promised severance settled in August 2025. That’s two-and-a-half years for a couple of trivial, open-and-shut cases.

The EU announced a $120 million fine in December 2025. And Elon lost a jury verdict last month regarding some light fraud he committed when acquiring the company. All the rest of the regulatory investigations (and there were several) were erased by the incoming Trump administration. They dragged on long enough for Elon to purchase the Executive Branch. He spent something like $278 million on the election. If Kamala had won, he likely would’ve been on the hook for much more than that in fines. (What do you call an electoral system like this? The Aristocrats!)

Those are… abominably slow processes. None of the investigations were particularly complicated. Elon was told (basically. I’m paraphrasing here) “hey boss, this is the law.” Elon replied, “lol no. That isn’t fun. We’re just not going to do that.” Elon’s lawyers said “jfc, fine, I quit. Here’s the number for the whistleblower hotline, y’all.” I remain generally Abundance-skeptical, but looking at how long these cut-and-dry investigations took is enough to Abundance-pill me at least a little bit.

The thing that I find particularly interesting here is the arbitrage opportunity it creates. Let’s just consider the employment lawsuits: Elon just decided not to pay what he contractually owed. The sum total of the two lawsuits (what they asserted he owed — the final settlement numbers are non-public) was around $628 million. By refusing to to honor his contracts, Elon Musk effectively had an interest-free loan from his counterparties for two and a half years. He could pay them right away, or he could hold onto the money and pay them later. In the intervening years, he was free to invest that money, make other deals with that money, and earn interest on that money. His counterparties were prevented from investing/making deals/earning interest/exchanging that money for goods and services.

That’s a tremendous financial deal. If you are rich enough, why would you ever honor your contracts in a timely fashion? Why not roll the dice on lawyering it out? Worst-case scenario, you get a multi-year interest-free loan. Best-case, your opponent settles at a discount, or outright loses on a weird technicality.

It’s only a tremendous deal, though, if you don’t have to account for the legal fees. Lawyers are expensive, and they generally charge by the hour. Elon’s lawyers are also expensive, but the trick is that he pays them by the year. They get paid whether they’re filing frivolous lawsuits or not. That’s a sunk cost, already on the books. So long as he’s paying them, he might as well keep them busy.

So that’s the jaded lesson I’m taking from this incorrect prediction: the legal system and the regulatory state move slowly, particularly when well-resourced actors are invested in slowing them down.

This is all fixable, but not without reforms to the legal system that won’t see the light of day anytime soon. For the near-term, this strikes me as a durable and terrible status quo feature. Never underestimate how much the people who pay their lawyers by the year can get away with. The asymmetry between their access-to-lawyers and everyone else’s creates arbitrage opportunities that they predictably exploit.

Some Personal News: I’ve been an avid reader Flaming Hydra since its inception in late 2020. Flaming Hydra is one of the few bright spots in our ever-dimming media landscape. A bunch of great writers decided “fuck it. We’re going to form our own collective. Everyone will contribute one essay per month, and we’ll charge a $3/month subscription fee. There is an audience for great writing, this doesn’t have to be complicated.”

The Flaming Hydra crew has expanded their ranks and… well… I’m one of the new contributors! I am going to try to be chill about this. I will not manage to be chill about it. This is so fucking cool.

They published my first essay, “A New Guy For Us All to Be Mad At” last night. It’s about the history of PayPal, and the sheer incompetence of young Musk and Thiel.

The main character of the essay is Bill Harris, who you have probably never heard of. Harris was hired to be CEO of young Elon’s X-dot-com (the dotcom-era payment company, not the present-day child porn/white nationalist content creation machine). As the lone “adult in the room,” he was annoyingly competent at his job. He engineered the merger of Elon’s X and Peter Thiel/Reid Hoffman/David Sacks’s Confinity. That landed them a $100 million Series C funding round, just days before the dotcom bubble well and truly popped. Without Bill Harris, the deal collapses, the funding round doesn’t come through, and both companies become casualties of the dotcom crash.

Here’s a little thought experiment: imagine you have a time machine, and it is your job to find the smallest historical event that, by changing it, reorders the course of human events for the better. My current candidate would be “go back to March 2000 and give Bill Harris a mild case of the flu.” Nothing too extreme. Just, like, a week or so on Nyquil. We’ve all been there.

That’s a parallel world where Bill Harris is stuck phoning it in at work, and Musk and Thiel have to be normal for a week or else their companies collapse (difficulty level: impossible). Here’s an accurate rendering of the result:

Dammit, Bill Harris. We could have had it all.

You can read the whole essay here. And please subscribe to Flaming Hydra.

A few other links and notes from the week:

(1) I felt a bit of honest-to-goodness hope reading Julia Angwin’s column this week. (Check out John Herrman, too.) Meta just lost two court cases about the impacts of its design features on social media users. The cost of those cases is barely a rounding error for Meta, but it also represents a potentially serious change to the status quo. Section 230 has provided blanket protections to the tech monopolies. The blanket is fraying.

I’m generally of the opinion that we are better off with Section 230 than without it. But, also, Silicon Valley has an iron grip on basically two branches of the government right now. And things are getting worse all over the place. The courts are starting to treat Meta like Philip Morris. That’s something to work with. I’ll take it.

(2) OpenAI is shutting down Sora. Lol. LMAO even. I’m not ready to declare this a turning point for the AI bubble (though it is still giving Enron, y’all). But OpenAI itself is looking more and more like a sinking ship. Disney’s $1 billion investment is toast. OpenAI’s near-term finances run on futurity. It only works if it seems inevitable.

On a related note, I have started gathering threads for a piece with the working title “Dario-Wario." Dario Amodei is pretty clearly replacing Sam Altman as the standard bearer for the generative AI revolution. So I’ve been trying to work out exactly what to make of Amodei. He is, I think, clearly less-bad than Altman. But that doesn’t mean he is necessarily good or trustworthy.

More on that soon(ish).

(3) The Giving Pledge has become a punchline. [Sean Parker in The Social Network voice]: “giving away trillions of dollars isn’t cool. You know what’s cool? Claiming you will give away trillions of dollars, basking in the credit for years, and then just… keeping the money.” Who could have ever seen this, the most obvious turn of events, coming?

We’re going to have to tax the hell out of the billionaires. Not taxing the billionaires was a mistake. Believing that the billionaires would collectively, voluntarily give their wealth to good causes was a mistake. There are a handful of exceptions (MacKenzie Scott deserves at least a few statues when the histories of this era get written), and/but all of the good billionaires would, by definition, approve of taxing the hell out of the billionaires.

(4) I’ve been reading Ben Recht’s book, The Irrational Decision: How We Gave Computers the Power to Choose for Us. It is excellent. I’m not sure if I’ll write a full review, but you should really put it on your reading list.

I’ve also been reading Francis Spufford’s new novel, Nonesuch. It is such a goddamn treat. Apologies to all other novelists publishing books in 2026. Spufford is going to be my favorite novel of the year. You’re all competing for second place.

I read Project Hail Mary over the weekend. I haven’t seen the film yet. I… viscerally hated the book. I might write something longer about it. I don’t want to yuck anyone’s yum — the book just belongs to a genre of whiz-bang science solves everything fiction that, it turns out, is extremely not my cup of tea. But I feel the urge to think a little harder in public about why it is so not my cup of tea. We’ll have to see.

That’s all from me. Thanks for reading. Have a nice weekend.

-DK