1. Great article by Sebastian Jensen, “The illusory difference between sincerity and irony.”
When people are being sincere, they do what reflects their thoughts and emotional state. Getting mad when the sports team loses. Claiming to enjoy foods that you enjoy. This opens people up to being vulnerable, being attacked for things that are linked to their identity. So people have started being ironic: doing things that do not reflect the self.
I can relate to this. Irony puts up a shield and helps to emotionally distance oneself from the unyielding nature of reality or an otherwise unfavorable outcome, as a coping mechanism. But satire, which is related to irony, is a powerful rhetorical tool, so irony shouldn’t be dispensed with altogether.
2. The AI bubble is 17 times the size of the dot-com frenzy and four times subprime.
This is a terrible article. Predictions of AI being a bubble are of zero actionable value. No one has any way of knowing if it’s a bubble, and if so, when it will pop, if it ever does.
He blames low interest rates, “Artificially low interest rates have stimulated investment into AI that has hit scaling limits,” but interest rates have surged since 2022 to their highest levels in decades. He cannot even get the basic facts right, which kills his credibility at the start.
This also undermines the common belief that high interest rates are always bad for asset prices. For venture capitalists chasing 40-100% annual returns in the hottest AI startups, the difference between borrowing at 1% or 5% is irrelevant and dwarfed by the expected upside. A similar pattern emerged in the ’80s and late ’90s, when high interest rates coincided with even higher valuations for publicly traded tech companies than seen today.
This means a much longer effort at reflation, a bit like what we saw in the early 1990s, after the S&L crisis, and likely special measures as well, as the Trump administration seeks to devalue the US$ in an effort to onshore jobs,” he says.
In an attempt to paint a negative picture of impending crisis, he gives examples of 2001 and 1991, which were also among the mildest recessions ever. The US stock market and economy would go on to boom in 1995, just a few years after the S&L crisis. If there is a job that AI needs to kill off, it’s these overpaid and useless analysts.
3. The AI-warning flop
I predicted here and here that Eliezer Yudkowsky and Nate Soares’ book If Anyone Builds It, Everyone Dies would be a flop. Sure enough, a month after its release on September 16, 2025 with much fanfare online, there are scarcely any mentions of the book on Twitter, indicating it has had little to no impact on AI discourse. Despite a large marketing budget and converge by such major media outlets as The New York Times and Bloomberg, and astroturfing online leading up to and shortly after its release such as podcasts and Reddit discussion, it was only able to secure a #7 debut on the NYTs bestseller list for the category “Combined Print & E-Book Nonfiction”. The number one book that week? POEMS & PRAYERS by Matthew McConaughey. Yes, a poetry book. So we’re not exactly talking a high bar here.
It dropped off the bestseller list for good in its second week. How many total sales does this translate into? Maybe a couple thousand optimistically. The publisher should have consulted me. I would have waived my fee to tell them to not to publish the book. The authors of course pocket the advance, so it was worthwhile for them even, with any future royalties deducted against it. But the publisher thought that this book would make waves in AI discourse, and it amounted to no more than a ripple.
Every debate about AI seems to end up back where it started. No one can really agree what it means for a machine to ‘think’ or what constitutes actual genuine intelligence. But LLMs do a reasonably good job of it, however this is defined. They are intelligent enough for the tasks required of them. It’s hard to make a case for AI destroying the world when the public sees AI as just a smarter version of Google.
4. Today in stupid: Autism Is Not a Single Condition and Has No Single Cause, Scientists Conclude.
The analysis, published last week in the journal Nature, showed that children diagnosed before the age of 6 were more likely to have behavioral difficulties—such as problems with social interaction—from an early age. In contrast, those diagnosed after the age of 10 were more likely to experience social and behavioral difficulties during adolescence. They also had a greater predisposition to mental health conditions, such as depression.
So earlier onset symptoms is correlated with an earlier age of diagnosis? Who would have guessed. If this is what passes for publication in a prestigious journal like Nature, then this already lowers my already low opinion of the field.
5. Centralization requires perfect foresight :
As we saw during Covid, some of the ‘smartest’ or technocratic countries had among the worst outcomes, such as repeat ineffective lockdowns, multiple waves of relapses, stunted economic recoveries, and protests and other unrest. Covid showed the downside of centralized government, as citizens of Germany, China, Italy etc. were unilaterally subjected to onerous restrictions, which in hindsight were ineffective. At least in the US, the restrictions were enforced more on a local level than nationally, so people had the discretion, in theory, to move to locations with fewer restrictions.
Such a decentralized approach also made it easier to test which policies worked or didn’t work. Also, such restrictions if excessive would be unpopular politically, discouraging overreach. Such mitigating factors are absent in more totalitarian systems. Curtis Yarvin’s centralized approach to Covid presupposed the conclusion that the centralized policy approach was also the optimal one. This cannot be taken for granted. Yes, with perfect foresight a benevolent dictator can craft optimal policies that makes society better off. But Covid clearly showed this to be false when dealing with uncertainty.