DanBot

I enjoyed reading this retrospective analysis of the 1990s-era predictions of Mary Meeker, one of the most famous internet-bull stock analysts from the dotcom era. Meeker was strongly associated with the dotcom bubble, and after the bubble burst she was seen as a poster child for irrational internet cheerleading. Yet 30 years later, this analysis by Dylan Litman (which I found via Bloomberg's The Forecast newsletter) finds that the vast majority of Meeker's at-the-time-crazy predictions ended up either mostly panning out or even being conservative in retrospect. One of her ambitious predictions was that global internet adoption would reach 7% by 2010; the actual figure turned out to be 29%. In the end, Litman found that only a small fraction of her predictions were unambiguously wrong. I don't have empirics, but my sense is the same about stock analysis and business commentary during the period of the smartphone. Smartphone bulls, and especially those bullish on the iPhone, understood their future much more acutely than did the bears. I'm confident a similar dynamic will obtain in AI. Even the bulls (I am one) are probably profoundly underestimating the size of what will come. There is much talk of, and much skepticism about, the size of the capital expenditure on data centers and related infrastructure. I don't like to speculate about such things in general, but a few observations: 1. Capital expenditure is a leading indicator; data centers take years to build, so capex reported today often does not bear fruit for quarters to come. You will see huge figures quoted today, and the size of these figures will set the vibes in AI business commentary. Yet you will not actually feel the effects of that spending as manifest in AI capabilities for quite some time. 2. To date, I believe all publicly available frontier models were trained on Hopper-class chips, the ones that were so popular in 2023 and 2024. Soon models trained on Blackwell and similar-class chips will be released. 3. In the coming year, OpenAI Stargate, Anthropic/AWS, xAI, Microsoft, Google, and others will be bringing gigawatts of capacity on line. In some cases, individual sites will significantly outmatch *all* the compute that frontier lab had available to it before this year. 4. We have not yet seen AI models trained and served on this new stack. I feel good about what these improvements alone can deliver in terms of model performance. 5. The AI industry is seeing strong adoption, though the stickiness of that adoption (both to any one provider, and to generative AI tools in general) remains to be seen. AI coding agents, a new form factor for AI beyond the chatbot, have barely begun to diffuse. These systems expand the utility of AI into a vast range of new dimensions. We have barely--and I mean less than a single-digit percentage point--scratched the surface. 6. All accounts indicate that the best AI systems today are themselves speeding up development progress within the labs; much of what a frontier AI lab does is software engineering, so the productivity benefits they see from coding agents in particular are very, very high. 7. It is probably the case that positive-feedback loops where AI systems build the next generation of themselves have begun. They will get tighter and larger in the coming year, possibly much so as the models begin to be able to automate the day-to-day work of frontier lab research staff. This is widely discussed in the industry. It is not science fiction. It appears likelier than not that it will at least begin within the next year. No one really knows what AI designing AI will mean, but it probably isn't *bearish for AI capabilities.* 8. Thus you should think of the frontier labs in this way: seeing huge revenue growth and future demand, on the cusp of bringing a new generation of infrastructure fully online, hot off the heels of a generation of frontier coding agents which to many have been shockingly capable, and preparing to scale their workforces from 1000s of human staff to potential 100,000+ human-AI hybrid workforces. 9. It may seem like you cannot believe your eyes. The capex numbers really are mind-boggling. The scale of construction by 2027 implied by current capex is astonishing. And yet look at the pattern of facts above. How compelling does a person yelling "stochastic parrot!!!" over and over and over sound in comparison to this basic set of facts, which isn't even close to the most compelling case a bullish person could make? 10. At some point, the rate of AI capex growth must slow. Nothing grows faster than GDP for ever, as my friend Oren Cass once said. Yet there is a compelling case to be made--indeed it has been my operating assumption since the fall of 2024--that despite all the ambition and the despite the scale of investment, the US will in fact be under-provisioned on semiconductor fabs and AI computing infrastructure in the late 2020s. Demand may well still exceed supply. I hate "bubble talk." I find it so boring. Will there be a dip here or there in the capex? Sure, maybe. But quite possibly not. The story of 2030 implied by the set of facts above suggests that there are *much* more interesting questions all of us--policy people, investors, journalists--should be asking. Almost none are. In the end I don't really care about what people say about my predictions today, or in 2027. I am quite confident that in 2055 we will look back on the AI bulls and say, "how bearish they really were!" FWIW I never felt similarly about crypto, though I always assumed it would do just fine.

the "chinese century" is a reddit ideology. having a bunch of LED's on your buildings and lots of solar panels doesn't make it your century sorry. it's an American century, same as the next 9

“China is building ‘full-stack’ defense-innovation cities" defenseone.com/technology/202… Our latest @DefenseOne with @BluePathLabs "While the U.S. struggles to add rare-earth factories and drone-test ranges, Beijing is creating them in clusters...The result is an industrial powerhouse that other Chinese cities are working to emulate—and that U.S. firms are unable to match."

I used to teach, mentor and tutor black people to help them pass IT certifications in my early 20s. I’d provide them with the latest training resources and pass on job referrals to them from recruiters I had worked with in the past. It was a field that didn’t require a lot of money to get into and provided great room for growth and stability for people who didn’t come from the best backgrounds. I spent a lot of my personal time doing that because I was fortunate to be doing well so early in life and wanted the same for others who had been way less fortunate than me. Today in 2026 I’d no longer feel comfortable doing this because the field has been over saturated with nepotistic “legal” immigrants. A bunch of the recruiters barely speak English and they offer fake jobs. The pay doesn’t scale the way it used to and the treatment of ppl in contract positions has degraded significantly. This was a very affordable and legal path to success for a lot of black ppl who came from rougher backgrounds but had some talent. It’s not entirely gone but I don’t feel comfortable encouraging people to sign up for it. Black Americans should absolutely be the most concerned demographic about immigration. It’s time to stand up for yourselves and your one and only home.

Europeans are so naive. If Europe breaks from the USA, the USA could very easily decide that the only way to counterbalance China is to break Russia off from China. How? By arming the Russians and letting them have Europe. Russia then becomes a credible rival to China, and Russia and China fight it out for the next 500 years while we Americans just watch from the Western Hemisphere. NATO was very beneficial to the USA when it was needed to balance against the Soviets in a bipolar world, when the USA and the Soviets were the only game in town. But with the rise of China, a potential tripolar world of the USA, Russia+Europe, and China is now arguably more stable than the bipolar world of USA+Europe vs China+Russia. The only place that really loses from this is Europe, which must be conquered. The only people who really really need NATO are the Europeans. Listen to the Finnish President talk. Even in his worse case scenario, the USA is still giving him the passwords to his planes. He can’t even envision a world where the USA says “fuck that,” leaves NATO, lifts sanctions on Russia, and starts shipping the Russkies weapons. But if Europe cozies up with China, why wouldn’t we do that? You think we’re going to fight Europe, China, and Russia all at once? No, we’re going to pick one to ally with. Russia shares a huge border with China. China is an immense threat to Russia. Way more of a threat than either China or Russia are to the USA. The only reason Russia is allied with China at all is because of how powerful the USA is right now, and how weak Russia is. And Russia is only weak because we have refused to let Russia get more powerful by eating Europe. Geographically and historically, a US-Russia alliance against China would be way way more natural than the current situation. Europeans need to wake up. This isn’t a time for silly games and childish fantasies. I love Europe, I love Europeans, I want you all to live happy and peaceful lives. But you haven’t considered at all what it would mean to have the USA as an enemy.

/3 shortening the supply lines with their build up and lock downs in Xinjiang. Vast factories at the very western edge of their border.

/2 Want to note, the plains of Kazakhstan are easily passable. Go through the Dzungarian Gate, where armies for centuries have crossed. Then vast plains. Yes you have to go through part of Russia, including Stalingrad, to get through Ukraine. The Chinese have spent years

Want to gently suggest that at some point, it makes more sense for China to invade Europe in WWII-style fashion. Nothing personal. Simply the result of too much excess capacity: Europe is simply available if one crosses the plains of Kazakhstan, into Ukraine, and then Westward.

I want to stress this, Europe ***does not have a middle class anymore*** as such China will dump its homogenous goods and its low cost cars into Europe with its own chinese workers, Chinese companies, SOE’s etc (Belt and road has a euro corridor) The chinese good cars (their midline and high lines are amazing) will go to Asia, where they can afford them and where those countries can soak up the surplus human capital. Then, just like with Germany’s “we went too far against nuclear”, Europe politicians in 20 years will say “we went too far with favoring China” and the whole cycle starts again.

Greenland is a sideshow. But there is signal. You should be asking… why is the admin willing to piss off the Europeans to this extent over something that while strategically valuable is far from a core priority? The answer is that the Trump admin knows something you don’t. They are going to be pissing off europeans much more aggressively over the next 12 -18 months anyways. Because they plan to enforce the transshipment clause in all those trade deals to force Europe to shut out China. And that is going to throw Europes economy into a tailspin. So what the admin is now doing is preemptively showing it doesn’t give a flying fuck about the “rules based order” If you don’t want Greenland in the hands of the Americans you clearly don’t take the long term threat from China and Russia seriously. And if that’s the case then fuck you. We don’t need you, Europe. And they are going to do a hundred things to show they mean it. Because remember what I said a year ago. Bessent wants to salvage the system. But if that fails the backup is the Vance doctrine. And that entails walking away and letting it all burn. When you start with a lens of strategic decoupling everything makes a lot more sense. 🫡

possibly my favorite meme

@Ryanair Should I buy Ryan Air and put someone whose actual name is Ryan in charge?

Soon kids will be doing a term in the French Foreign legion in order to make their Ivy applications stand out. It's an arms race, starting a fake non-profit and playing the violin won't cut it anymore. You need to be seeing combat in Mali.

Pro-question: which Chinese hacking syndicate uses one of these US cybersec firms to obfuscate their activities even though the firm and their clients have been targeted by this hacking syndicate?

If Iran also falls, than perhaps all of the governments of the world are weaker than previously thought. From there eyes shift to North Korea and eventually China. All the things the Bush Administration dreamed of, are now coming to pass with limited US involvement.

today is january 13, 2026, reminder that most american OEMs (aerospace, defense, automotive, etc) have basically zero supply chain visibility and things at the third tier are basically all chinese.

Every year during the Kentucky Music Educators Association (KMEA) conference in Louisville, hundreds or thousands of high school choir students spontaneously gather in the atrium of the Hyatt Regency Hotel to sing the National Anthem from the balconies, creating a powerful tradition that amplifies their voices and has been happening since the 1980s.

@planefag Something to think about. If you look at the NY Times front pages in the 4 days~ before Pearl Harbor, there are headlines asking "Where is the Japanese Fleet?". Their movements were obvious. Our ability to respond was not.

Sweet jesus people underestimate the level of recon we engage in

@planefag In surprise attacks like these, worth noting it's never about completely sneaking past the enemy. It's about doing a good enough job, and hopefully the enemy (which definitely noticed your movements) can't push it up the chain of command fast enough for a response.

Trite observation maybe, but reading about Portuguese and later Dutch merchant traders traversing the uncharted open seas and enduring unspeakable peril to reach the Orient (the first sanctioned Dutch expedition to Java in 1596 lost 150 out of 250 men and resulted in a minor colonial war) just for a few pots of black pepper, literally just a few pots, and upon its delivery back to Holland was so alluring that procuring more black pepper became the central focus of the nation's treasure and manpower, and on the back of which an entire centuries-long shipping empire was built, as I casually, unthinkingly, in the most mundane and ungrateful way possible, grind a few dashes of same-said pepper over my scrambled eggs, induces the strange and I suspect wholly modern sensation of the enormous wealth and extravagance of ordinary life, and how laughably trivial so much of it is, but also for which I nonetheless give thanks.