Your local bodega owner is safe for now
And: The AI talent wars are getting expensive
Welcome to Cautious Optimism, a newsletter on tech, business, and power.
Happy Monday! I’m still chuckling at Anthropic’s experiment in making an AI model the head of a tiny store. It turns out that while AI can do a lot, issues like hallucinations and getting browbeaten by users into offering discounts mean that your local bodega owner is safe for now. Experiments like this are good fun, and show both the promise, and current limitations of our collective silicon brains.
Speaking of which, human authors are trying to find some way to push back against their work being ingested into AI models either for free or for tiny outlays. While many scribblers are trying the courts to press their case, they are also lobbying their publishers. To work! — Alex
📈 Trending Up: Guest worker programs? … Tandem Health … going out with a bang … data centrality … Cato Networks … Menlo Ventures … open-source AI … stablecoins … QSBS? …
📉 Trending Down: Wind power … national unity … Temu … F1 movie doubters … public accountability … prices in China … digital services taxes? … Microsoft’s home-grown chips …
Opening riffs
Economic signals: In South Korea, industrial production fell 2.9% in May (seasonally adjusted), compared to the preceding month. That was worse than expected. Retail sales in the nation were flat in the month. In Japan, industrial production rose 0.5% in May, a full 3% less than expected. In China, manufacturing activity contracted for the third straight month, but not by much. Non-manufacturing PMI, however, stayed in expansion territory while giving up a fraction of its growth. And last Friday we learned that domestic consumer spending contracted as consumer prices got a little bit hotter in May.
It’s a mixed collection of data, but mostly negative. To sum: The global economy is looking brittle, and without a great growth driver today.
American austerity: If you want to cut deficit spending, you can cut spending, raise taxes, or both. In the United States, we’re running our own playbook, cutting spending while making trillions of dollars in new borrowing capacity to extend tax cuts that would have tapered off automatically. The result? A funding bill that few like, including many of the President’s former allies. Elon Musk attacked the bill’s tax benefits for coal and increased friction for solar and other modern energy sources. Medicaid will get slashed. It’s a dog’s breakfast in polling terms, but looks like it will pass all the same.
Noahpinion has a good dive into the various energy levers in the bill, asking Would you rather have cheap energy, or stupid culture wars?
AI exemptions narrow: There’s a debate in Congress concerning whether its smarter to limit state-level AI rules to allow for a national playing field to develop. The idea — replete with industry backing — has support in the House and Senate, but getting enough GOP members aboard has taken some tinkering. From a ten year ban to a ten year pause in the Senate that would only be enforced by a loss of certain funding if states go ahead with their own plans to today, the latter but with a five year window. That’s not much protection, as the potentially lost carrot is merely a slice of one $500 million bucket. Cut that up by state and the pain for going solo is modest. And there are carveouts for things like child safety and the IP rights of musicians. Still, it seems that we’re reaching the middle here, and there’s good chance that state-level AI rules will remain a force, especially in states that can afford to fund their own broadband.
Maybe the prediction markets are good for something?
The AI talent wars are getting expensive
Reporting that Meta’s Mark Zuckerberg was getting deep into the AI weeds to ferret out the names of people truly executing groundbreaking research led to The List, a document of the folks that the social media giant’s burgeoning ‘superintelligence’ team may want to hire. Offers for folks on the list — and perhaps not — are high enough to shift gravity, though it appears that the much-vaunted $100 million packages are either incredibly rare, or simply a one-off that got overreported.
Changes to Meta’s AI efforts could constitute more than simply the staffing of a new AI team inside of its walls. The company, per the Times, at least considered backing away from its quasi-open source Llama model family. Instead, a more traditional closed-sourced set of products could be introduced.
Those are easier to monetize, mind. For every billion dollars worth of ARR that OpenAI and Anthropic accrete, they hack away at their burn. In short, closed-source AI today is a good way to make AI projects less costly in economic terms. Meta is spending ever more, and doesn’t have the same contra-cost lever. And even Zuck’s empire doesn’t have unlimited resources.
OpenAI has found its footing after ceding some talent to Meta. The company’s chief research officer Mark Chen wrote in an internal memo that he felt as “if someone has broken into our home and stolen something,” referring to Meta’s pilfering of its human talent. Expect Sam Altman’s company to fight back with upped comp packages of its own.
The way to make all the math math is to generate large, quickly growing, and gross margin-competent revenues from AI. OpenAI is doing well at at least the first two line items, even if we aren’t too aware of its gross margins as they stand today. OpenAI can therefore afford to up its pay to key folks; it grew from $3.7 billion worth of total revenue last year to a $10 billion worth of annual recurring revenue last month.
But still much, much larger. So much bigger in fact that its Q1 2025 net income was $16.6 billion. OpenAI will need to double, double again, and double again to reach about half of Meta’s total revenue scale.
But at the pace it’s growing, and with its complete focus on AI and not the multi-product collection that comprises Meta, in time the financial advantage that the social giant enjoys today should moderate. That fact helps us understand why Meta is going so hard now: For every month that Meta could have waited to try and poach critical talent from OpenAI, the AI company gets bigger, growing at a far greater clip than its own 16% growth it recorded to start 2025.
So, better now than later, and better now than never. Meta smashed the last few major technology cycles. It rode social to the moon, and then mobile to the stars. But now the world is changing yet again, and it doesn’t want to wind up as Microsoft did when it missed mobile. So, spend now, and hope that you win the race to AGI, or however close we’ll get in the next few years. The plan makes sense, it’s just going to be punishingly expensive.
Bear in mind that as we consider the rising cost of the humans building next-gen AI software, the hardware those systems will require is being built out at a pace that will either resolve to the best pre-investment of all time, or an albatross of epic proportions. I’m an Option A kinda guy, but admit that there is probably an investment ceiling, after which marginal dollars are less efficiently spent today on new datacenter capacity.
Do yourself a favor
Closing today, it feels odd that the person who correctly predicted the democratic primary in New York City was Peter Thiel, ages ago. From an email thread between himself, Mark Zuckerberg, Nick Clegg, Sheryl Sandberg, and Marc Andreessen:
Nick -- I certainly would not suggest that our policy should be to embrace Millennial attitudes unreflectively. I would be the last person to advocate for socialism. But when 70% of Millennials say they are pro-socialist, we need to do better than simply dismiss them by saying that they are stupid or entitled or brainwashed; we should try and understand why. And, from the perspective of a broken generational compact, there seems to be a pretty straightforward answer to me, namely, that when one has too much student debt or if housing is too unaffordable, then one will have negative capital for a long time and/or find it very hard to start accumulating capital in the form of real estate; and if one has no stake in the capitalist system, then one may well turn against it.
Given that analysis, it is not a massive surprise that the way Zohran Mamdani won his recent race was by getting a huge number of younger people to vote for him.
As a capitalist, I worry about elements of the Mamdani platform. But I wonder what we’re offering as an alternative. Cuomo was a non-starter, but abstracting from one city’s peculiarities I think that regressive tax moves, investing the national dollar in coal while curtailing green energy production, protectionism, centrally-set prices, government-owned industrial capacity, punitive moves on student debt and the like are not the way to win the youth back over to free markets. I think that the people who claim the mantle of capitalism are often in fact fomenting their own losses at the ballot box. And blaming the youth, per Thiel, won’t get you very far.