Strong piece, though I'd push back slightly on one thing: I don't think the format is the root cause so much as a convenient amplifier. The consumer tendencies you're describing the entitlement, the harassment, the hype cycle those exist in film fandom, in sports, in music. The not-E3 presser makes it louder and more concentrated, sure, but I'm skeptical that a better format produces meaningfully better behavior from the same audience. That said, the point about smaller games being used as spacer material between megaton reveals is one I hadn't quite articulated to myself before reading this. That's a real structural problem. An indie with an honest trailer getting buried not because the audience dislikes it but because it exists in the wrong slot that's a format failure, not just a culture failure.

The trap in all this "permanent underclass" conversation is that it keeps framing the problem as something new. It isn't. What AI does is accelerate a concentration that has been underway since the 1980s with the financialization of the economy. The difference now is speed and generality: this time there is no cognitive sector left to retreat to. The point nobody wants to say plainly is that the problem is not technological, it is one of ownership. Whoever holds the models, the training data, and the infrastructure captures the productivity gains entirely. The worker who was replaced has no stake in the asset that replaced them. That is the real discontinuity from previous industrial revolutions: a factory needs workers to operate, a language model needs nobody once it is trained. The UBI debate is symptomatic of this confusion. UBI is a redistributive response to an ownership problem. It is the equivalent of paying unemployment benefits to someone who was expropriated without compensation. It solves subsistence in the short term, it does not solve the intergenerational wealth accumulation that will calcify social hierarchies permanently. What should be at the center of the debate is co-ownership: sovereign AI funds, mandatory worker equity in companies that replace them, public licensing of models trained on public data. None of these mechanisms are technically difficult. They are politically impossible as long as capital remains concentrated where it is. For the Global South the picture is worse, because it is not just domestic unemployment: it is the destruction of cheap labor as a competitive argument. Angola, Bangladesh, Vietnam competed for industrial investment by offering low-cost work. That argument is over. Industrial AI does not need geography or low wages. What remains are natural resources and consumer markets, which are exactly the assets the Global North has always controlled through other means. The permanence of the underclass does not come from the technology itself. It comes from who arrives at the future without assets. It has always been that way. AI just made the timeline shorter and the scale larger.

They spent decades and millions in litigation defending these districts, the Supreme Court sided with them in 2024, and now they're already redrawing everything again to lock in a 7-0 sweep. The question isn't whether there will be a lawsuit — there already is. The question is: will it move fast enough to change anything before the August primaries? Probably not. And everyone knows it. The slowness of the courts is the strategy.

Whether it was deliberate, negligent, or just an extraordinarily unlucky coincidence, the outcome was exactly the same. And that's what makes this so uncomfortable to resolve. At some point, "we didn't mean it" stops being a defense and starts being an indictment of its own kind. How do you run a major operation in a country without anyone in the room knowing what May 18th means? The internal review found no conclusive evidence of malicious intent, yet some employees refused to hand over their phones. So we're stuck in this strange middle ground: a company that was either deeply cynical or deeply ignorant, with no way to tell which, and an apology that was, frankly, flawless in execution. Which almost makes it worse. The real question nobody seems to be asking is whether intent should even matter here. If the harm is real, does the reason change anything?

What this episode reveals is something deeper than a simple messaging problem. Framing everything as "bad messaging" is, to some extent, a way of softening what is actually happening. American defense officials themselves admitted, anonymously, that they had "spent the better part of two weeks reacting to the first announcement" and that they didn't know what the second one meant either. That isn't poor communication — it's the absence of a coherent decision-making process. The pattern here is one of foreign policy driven by impulse and personal grievance. The first withdrawal was announced after Chancellor Merz criticized how Washington was handling negotiations with Iran, and the second announcement came when Trump questioned Hegseth about why the deployment had been cancelled, saying the U.S. should not "treat Poland poorly" — apparently because he simply likes Nawrocki. In other words, the American military posture in Eastern Europe is being determined by political affinities and momentary dislikes, not by strategy. The problem with the "messaging" framing is that it assumes a well-defined policy exists and is simply being explained poorly. But allies are still trying to determine whether the administration is reducing its commitment to NATO overall, or simply reshaping it around governments Trump sees as more loyal. That uncertainty cannot be resolved with better communication. It stems from the policy itself being incoherent. The European response is telling. NATO chief Mark Rutte praised Trump's decision while simultaneously stressing that it would not change the push for Europeans to become less dependent on a single ally. The alliance's own leadership is publicly managing American unpredictability, preparing its members for a future in which the U.S. may not be there. That is an enormous signal, and it plays out regardless of whether Washington's messaging is clear or confused. Trump has even suggested he might pull the U.S. out of the very alliance it helped found after World War II, which places any one-off troop announcement within a context of permanently degraded credibility. Poland may celebrate today, but what Sikorski called "all's well that ends well" is really a normalization of a transactional and volatile relationship — one that Poland, given its geography, is less able to ignore than most allies. The question the article raises about poor messaging is real, but secondary. The core problem is structural.

The debate Bezos is missing, or avoiding, is about the difference between tax incidence and tax visibility. Federal income tax is the most visible part of the American tax burden precisely because it shows up as a single line on a pay stub. But visibility isn't the same as weight. A serious proposal to help struggling households would start with payroll taxes, which are regressive by design and cap out in a way that systematically favors high earners, or with consumption taxes, which take a larger share of income from people who spend most of what they make. There's also a fiscal arithmetic problem that the proposal quietly sidesteps. If the bottom half of earners accounts for only 3% of federal income tax revenue, eliminating that share doesn't free up meaningful money for those households — it just shifts the accounting. The nurse in Queens gets $1,000 a month back, but if the services that keep her neighborhood functional get underfunded as a result, the net gain is far less obvious than Bezos makes it sound. The smartest version of this argument would be a straight expansion of the Earned Income Tax Credit, which already functions as a negative income tax for the working poor and has decades of evidence behind it. Bezos doesn't mention it, probably because it already exists and doesn't need a billionaire to champion it.

Solid writeup. One thing worth adding on the Mullvad side: their decision to drop port forwarding in 2023 was actually a privacy move, not a cost cut. Port forwarding made it easier to correlate users over time. Most people complained about torrenting, but the tradeoff makes sense if your threat model is anything above "I just want Netflix from another country." Also worth mentioning that Mullvad Browser, built with the Tor Project, ships with fingerprinting resistance baked in. Pair it with the VPN and you're covering two attack surfaces most people completely ignore.

This is exactly the point nobody wants to say out loud. We've spent years celebrating the "thousands of contributors" narrative while quietly depending on a handful of exhausted volunteers to keep the whole thing from falling apart. AI didn't break the system. It just arrived at a house that was already on fire.

Honestly, this feels more like a strategy than an actual “improvement.” It starts with 5GB and then kind of nudges you to give your number to get back to 15GB… which says a lot. At the same time, I get the spam and fake account issue. I’m just not sure if this really solves it or just creates a different problem.

There is a dimension the title captures well but deserves to be unpacked further: the "paying twice" is not only financial, it is reputational and cultural. Companies that conducted mass layoffs wielding AI as justification now face a talent market with a long memory. The most skilled professionals, precisely those these organizations will need to rehire, watched what happened and are pricing in the risk of working for employers who instrumentalize technology to justify decisions that were primarily about cost. The CTO who needs to rebuild an engineering team in 2026 will find that trust cannot be reconstructed with a competitive salary package. That is the second installment of the debt, and perhaps the most expensive one of all.

Jensen Huang was originally left off the list, then Trump called him personally and he flew to Alaska to board Air Force One. That last-minute scramble tells you everything: AI chips are now as central to US-China diplomacy as tariffs and Taiwan. The real question is whether Huang's seat at the table will loosen export restrictions on Nvidia's most advanced chips, or whether this is just optics at 30,000 feet. What do you think China will ask for in return?

When Trump and Putin say the same thing on the same weekend, it's worth asking: who is negotiating with whom, and in whose favor? Was the 3-day ceasefire a real signal or just a symbolic pause for the cameras? Is Ukraine sitting at the table as a sovereign nation or as a piece of the deal? Share your take.

I get the feeling the market is kind of “forcing” an optimistic logic. Like, semiconductors are doing well, so everything must be fine. But recent history shows that heavy concentration in a few sectors can inflate the index while hiding weakness underneath.