I want to briefly explain why I joined Abhijit Banerjee, Peter Diamond, Esther Duflo, Paul Krugman and Joe Stiglitz in signing the letter on the California billionaire tax: In principle, I am not convinced that permanent wealth taxes would be necessary if the tax-transfer system were designed optimally. But the US tax system is very far from optimal, and has been for decades. As I have documented in my research (for example, here: labor income is taxed much more heavily than capital income. This asymmetry creates two distinct problems. First, it makes the tax system highly regressive at the top. The very rich, who receive much of their income from capital or can use accounting tricks to reclassify their income as capital income, pay remarkably little in taxes. For example, a business owner who runs their own company should receive a significant part of their income as labor earnings for their work as CEO. Instead, they can take their compensation in stock and borrow against those holdings to finance whatever consumption they desire, minimizing their tax obligations. Even their heirs may avoid paying these taxes. Second, the asymmetry distorts automation decisions: it effectively subsidizes machinery and AI relative to hiring workers (for example, here: These distortions have allowed a small number of people to amass vast fortunes without paying their fair share of taxes, and have fueled excessive automation. The resulting inequality is a problem in its own right. It is all the more dangerous today because our institutions have become fragile, allowing the very wealthy to exert growing control over the political process. A temporary wealth tax can therefore be justified on three grounds: (1) it partially reverses the effects of more than two decades of tax avoidance by the very wealthy; (2) it acts as a brake on their growing dominance over the political process; and (3) it may pave the way for more comprehensive tax reform at the federal level. The California billionaire tax is not perfect. For example, a federal tax would lessen risks related to capital flight, and removing the rigid earmarking of the revenues for specific purposes would enable the proceeds to reduce the national debt. Nevertheless, with few other options on the table, I believe the California proposal deserves support.

@tylermacro10 From SpaceX's actual 2025 revenue of $18.67B, reaching ~$1T by 2030 implies roughly 122% CAGR over 5 years. To ~$3.5T around 2033 is about 92% CAGR over 8 years. The $100B ARR by end-2026 would require a sharp near-term acceleration from that base.

1/ One thing about commodities that increasingly frustrates me is the constant government hyperventilation whenever energy gets tight and prices rise disproportionately. It’s as if cheap energy is a basic human right. Well, it isn’t. 2/ The entire point of a commodity price is to balance supply and demand — every day, in real time. If we don’t have enough diesel, the price must rise until somebody at the margin consumes less, somebody produces more, or another barrel gets pulled in from somewhere else. The price is the invisible hand doing the heavy lifting. Suppress that signal and you don’t solve the shortage. You make the adjustment mechanism worse. Having such a signal available worldwide is the magic: a market price is an extraordinarily efficient mechanism for compressing dispersed information and coordinating behaviour without anybody needing to possess all that information. A diesel price incorporates, in one number, millions of changing facts: refinery outages, crude availability, freight cost, inventories, weather, demand, alternative fuels, storage economics, credit, geopolitical risk, and what every buyer and seller is willing to do at the margin. 3/ Yet whenever the signal becomes politically uncomfortable, governments attack the signal instead of the shortage. The US considers diesel export restrictions. Poland proposes a windfall tax and fuel-price intervention. Britain introduced its windfall tax in 2022. The EU followed with its “solidarity contribution.” Different instruments. Same reflex: prices are high, therefore punish the price mechanism or the companies responding to it. 4/ Take a US diesel export ban or quota. Sure, it could reduce US diesel prices in the short run. Then the second-order effects begin. If US diesel inventories build because refiners cannot export the surplus, refining economics deteriorate and refiners have an incentive to reduce runs. But a refinery doesn’t produce only diesel. It produces gasoline, jet, naphtha, LPG and other products from the same crude barrel. Congratulations: you “fixed” diesel and may have tightened something else. Meanwhile Brazilian farmers get the bill. 5/ Poland’s conservative government is now playing the same Marxist nonsense game. It already capped fuel prices this summer and is now proposing a windfall tax. How does confiscating refining profits create one additional barrel of refining capacity? It doesn’t. If anything, you are taxing precisely the economic signal telling refiners and entrepreneurs: WE NEED MORE CAPACITY. 6/ And this is the particularly absurd part in Europe. For decades Europe made building and operating refining capacity progressively less attractive while simultaneously encouraging diesel consumption. Then a global refining shock arrives, diesel cracks explode, and politicians discover — astonishingly — that Europe doesn’t have enough diesel. Their solution? Punish the remaining refiners for the shortage. 7/ How about doing the opposite? Give the industry a fast-track permitting regime for major refinery expansions. Make strategic capacity an explicit energy-security objective. If necessary, provide cheap, long-duration financing tied directly to incremental capacity. Poland could also build a much larger strategic crude reserve, available to domestic refiners during genuine supply emergencies under transparent rules, with any subsidised crude value contractually passed through to consumers. 8/ None of that lowers diesel prices tomorrow. That’s precisely the point. You cannot legislate a refinery into existence tomorrow. But you can make damn sure the capacity exists when the next shock arrives. The OECD oil market has an extraordinary ability to correct shortages and gluts through substitution, arbitrage, inventories, changing refinery runs and ultimately new investment. But only if politicians allow prices to do the work…! Not difficult to understand. @donaldtusk @sikorskiradek

Claude has discovered a previously unknown enzyme system hidden in the DNA of bacteriophages. Beside the enzyme’s gene sits a long array of repeating DNA—a structure that looks somewhat similar to CRISPR. We don’t yet understand what this system does, but only a handful of known systems share its features, and all of them are able to cut, copy, and paste DNA. Historically, the discovery of such programmable systems has helped revolutionize medicine. CRISPR, for instance, is now the foundation of genetic medicines. But it will take much more work to learn what this system does, and whether it can be put to similar use. Read more:

It has become fashionable to quote our token index as support for a bearish view on the AI trade. We have pushed back gently a few times, because nothing here is definitive and people should come to their own conclusions. It now seems worth saying a little more. In June we clarified what our LLM Token Index measures: what API users in our sample actually paid per million tokens. A usage-weighted price. Not token volume, not total spend. We noted it can be read, loosely, as revealed willingness to pay for frontier intelligence. We should have emphasized the conditions. The binding one is that the intelligence content of a token stays stable. Over a few weeks that is defensible. Across a couple of release cycles it clearly is not. While a token is simply the wrong unit for intelligence, measuring capability itself without bias is also extremely hard. [Btw we will struggle with this measurement problem all over the place as AI proliferates delivering non-market economic value.] The upshot is that the message from June landed. Maybe a little too well, because it created a new misread: that a falling index is necessarily bearish for the AI trade, since if token prices fall, model-layer margins must follow. Two things are being conflated. The deflation in our sample is real, and it is recent. It dates from the end of May. The index peaked at $2.07 on May 28 and sits at $1.00 as of Sept 21, down 52%. But the same index rose 67% from January into that peak, and few read the rise as bullish for lab margins. It is not bearish now. A usage-weighted price moves with the mix, in both directions. What the mix actually says: more work, at least within our sample, is being routed to cheap, fast models, and labs keep shipping more mid-tier variants. Most everyday tasks never needed a frontier model. That is partial equilibrium for the users we track, and our methodology note is explicit that this index alone cannot separate substitution from efficient agentic routing. Compute demand is a different question entirely, answered by different data series. Our H200 non-hyperscaler rental index has rerated through the summer: $2.87 average in June, $3.29 now, with a record $3.32 on Sept 19. B200 is $5.76, up 7% over the same stretch and 31% year to date. B300, our newest and thinnest series, is up 44% since inception in late April. H100 is off 7% from its August high, consistent with workloads migrating up the stack. Rents on the parts that are actually scarce are not signaling a demand stall. A world of mass agentic use is one where cheap tokens are nearly all of the count. Total token usage should keep growing far faster than the price is falling, with most of that growth coming from cheap, fast, and increasingly open models. The usage-weighted price can keep falling anyway. Our view, not a finding from this index: the highest value-add work still routes through frontier models, and that is where most of the economics will accrue. In any event, the implication for compute is more, not less. An astronomical number of tokens, most of them from cheap flash models, is what economy-wide AI proliferation should look like. It is not a demand stall.