What did Meta buy? A person? A dataset? A capability?

And NYC better become a smoking crater else some people are going to look silly

Welcome to Cautious Optimism, a newsletter on tech, business, and power.

Happy Thursday! Crypto price movement has been subdued in the last 24 hours, while total stablecoin supply is up $731 million in the last seven days, a gain of 0.3% to $252 billion.

It turns out money can buy friends. AI friends at least. Meta’s money-dumping efforts to poach AI talent from rivals like OpenAI is proving at least partially successful. Meta’s under-construction super-intelligence team has scooped three staffers from Altman’s Zurich office, the Journal reports, noting that the same trio worked at Google DeepMind before starting the OpenAI office in Switzerland. (There’s a movie about this?)

Elsewhere a bipartisan group of Senators has reintroduced the Open App Markets Act, which failed to become law back in 2022. Apple and Google hate the bill, because it would break their grip on application marketplaces on their iOS and Android mobile operating systems. If you want more purchase options, protected sideloading, third-party app stores, and less control over what can be built for your smartphone, it’s a bill you will probably like. (Read the full text here.) If you are part of the teams charged with keeping the two leading mobile operating systems secure and profitable, it’s less attractive. To work! — Alex

• 📈 Trending Up: Age verification in the UK … new home sales … Zhipu AI … AI gains … battery tech … Metaview … the domestic nuclear industry … births in South Korea … Chinese cars in Africa …

• 📉 Trending Down: Duh … personal privacy … AI gross margins … AI app makers? … rice prices in Japan … NATO squishiness …

What did Meta buy?

From one perspective, Scale AI is an incredibly successful startup. With $1.6 billion in raised capital, a valuation that soared to $13.8 billion, revenue growth to a reported $870 million last year, and an admittedly complicated but massive quasi-exit to Meta for $14 billion.

On the other hand, the company missed its revenue and profit goals last year. And before shedding large customers after linking up with Meta, it was already seeing some of its most famous customers move away. Bloomberg reported in the wake of the Meta-Scale deal that in the year or so leading up to the deal with the social media giant, OpenAI had already “determined that Scale was not the best fit for it because the AI developer needed more data expertise than Scale could provide.”

More, Inc. reports that Scale’s historical work for Google was, at times, sub-par.

The other perspective you could take on Scale — if you were feeling particularly cynical — is that it was aces at capitalizing on early genAI data-labelling demand, but may not have continued to grow as anticipated. The startup faces a burgeoning list of well-funded startups taking different approaches to the data labelling game and its related side quests.

So, what did Meta buy? A person? A dataset? A capability? I’m not entirely sure, but sometimes little cracks are simply cosmetic, while other times they are hints at something more structural.

I’ll bet you

Talk to folks who interface with LPs on the regular and you’ll learn that the folks who back venture capitalists don’t like to hear investments made described as bets. It’s too gambly a term for the risk-conscious LP world.

Some people are no fun at all.

Anyway, venture capitalists are gearing up to make new investments into PolyMarket and Kalshi, which allow users to invest in different outcomes from the real world: Where interest rates will move; how many times Elon will tweet; who wins a particular sports game; the price of cryptocurrencies in the future.

Wait, I can hear you shouting, some of those things are legitimate investing methods and thus shouldn’t be called bets. Yes, hedging is a standard activity. But taking a flier on, say, League of Legends results or whether or not Bruno Mars stays the most streamed monthly artist on Spotify are not risk-hedging wagers.

Anyway, the venture crew are gearing up to put more funds into Polymarket and Kalshi, and the sums are large:

• Kalshi just announced a $185 million at a $2 billion valuation.

• Polymarket is widely reported to be working on a $200 million round at a valuation north of $1 billion.

Two unicorn prediction markets? You love to see it. At least the competition between the two should keep prices low for users. I remain a bit skeptical at how accurate the two companies’ markets turn out to be — or it may be better to say better than existing methods of tracking market expectations for highly-expected outcomes. But the platforms also extend wagering into new domains where there might not be good price discovery today.

Hey, I play a lot of poker. I enjoy a flutter. Have fun!

Much ado about books

Briefly, we discussed the Anthropic suit brought by authors the other day. A judge found that while some of Anthropic’s use of copyright-protected books that it purchased was covered by fair use (a win for the AI company), its use of pirated books was not (a loss for the AI company). Since then, several other things have happened that are worth adding to our rolling IP-AI saga:

• Meta won a case against a number of well-known authors who felt that their IP was being abused by the AI giant. The judge, however, was more critical of the case brought than they were supportive of Meta’s arguments. Call this on a near-term win for the technology industry, and a potentially temporary loss for the scribblers.

• Microsoft is being sued by authors for alleged use of several hundred thousand pirated books to train an AI model, which the writers feel is against the rules.

• Getty is narrowing its case against Stability AI in the UK.

It will prove a very interesting situation if courts decide that books are fair use for AI training so long as the AI model company purchases them itself, and that use of pirated materials is not and must be compensated for. Such a world could enact huge penalties on AI companies for their past work that — may have — used pirate libraries like LibGen but less fines moving forward.

What would be better is if the very wealthy AI model companies offered fair compensation for use of protected materials that help make AI models tick. You know, the way business usually works.

• If Meta can spend so much on Scale, surely it can afford to pay for the raw materials that fuel its great AI engine?

• This is all very silly, and gross.

NYC better become a smoking crater else some people are going to look silly

I had a large segment planned for today going over what socialism means, and why or why not Zohran Mamdani deserves the label. Instead, we’re going to be brief as hell.

Roll with me:

• Defining socialism similar to defining other terms that are used by some as a compliment, and others as a pejorative. Get ten people in a room, and ask each to define feminism, for example. You’ll get ten different answers.

• Still, socialism can be fairly described as an economic system under which the means of production are collectively owned.

• The Democratic Socialists of America, for example, want “to collectively own the key economic drivers that dominate our lives, such as energy production and transportation.”

• Mamdani is a self-described democratic socialist, but his platform is incredibly light on plans to seize the mans of production. Instead, he imagines a number of public grocery stores which, by not having to “to pay rent or property taxes” can “reduce overhead and pass on savings to shoppers.” In short, he wants to use state subsidy to offer cheaper foodstuffs to citizens of the city.

  • Quelle horreur!

• At the same time, some of Mamdani’s proposals are very capitalist, including plans to cut “small business fines in half, speed up permitting and make online applications easier. He also wants to “increase funding for 1:1 small business support by 500 percent,” which is a bit dodgy in my view, but no one is perfect.

Sure, from my viewpoint, it would be better if Mamdani’s economic views more aligned with my own. But what’s staggering to me in the ~ discourse ~ about the Democratic nominee for mayor of New York City is how folks are pretending like he’s alone in wanting to meddle with business.

If such tinkering with the market earns on the scarlet S then what do we call this stuff:

• Extending subsidies for farmers growing crops that can be used in liquid fuel production, to the tune of tens of billions of dollars.

• Forcing businesses to change their practices, rewarding donors with political influence, and punishing companies that are construed to be enemies of the state.

• Getting in the middle of international M&A, handing POTUS control over a major American company, affording Trump the power while in office to effectively control the company, including setting a floor on the price it can charge for its products.

Influencing the market by pouring public money into certain business activities that are favored for political reasons? State meddling in the operations of for-profit companies? Direct executive control over industrial powers?

That sounds pretty socialist to me. And none of that stuff is coming from Mamdani. It’s coming from the second Trump administration, the same admin that some of the folks kvetching about Mamdani publicly support. Put that in your pipe and let it smolder.