Longreads + Open Thread

Longreads

• In Vice, Duncan Fyfe tells the story of how Sierra, the gaming company, was killed by an acquisition. That happens from time to time—it's hard to integrate a creative business into a more typical kind of company. But in this case, there was the added element that the acquiring company was cooking its books. This, at least, is a form of fraud with an exit strategy: the goal was apparently to use a good-looking mostly-fake business to pump up the value of the stock, in order to acquire a portfolio of actually-viable businesses. One notable theme, though one that may be exaggerated in hindsight, was how many little lies the fraudsters told: they played games with depreciation, executive departures, and the CEO buying an expensive bottle of wine he insisted he'd pay for himself, but ended up expensing to the company. You can usually see this kind of thing coming.
• John Luttig argues that the future of LLMs is closed-source. AI models are expensive to train, and while building one on spare hardware can be justified at low costs, it's harder to rationalize when costs run into the billions. It's probably true that the best models will be proprietary, and that you'll either pay to access them or wish you could. But it's also true that there will be a place for open-source for a long time: the biggest companies won't spend billions of dollars purely as a gesture of goodwill to developers, but might be willing to spend similar sums to weaken a competitor's economics.
• Bret Devereaux of the wonderful A Collection of Unmitigated Pedantry (ACOUP!) has a review of Alexander the Great's record as a military commander. The main thrust is that, yes, Alexander had a good track record, but also, that anyone who had inherited his situation would have had a big leg up over everyone else. It's a good reminder that history is contingent, but also that the most extreme historical figures tended to be both lucky and good at pushing their luck.
• Nilay Patel at The Verge has a sometimes hostile interview with Sundar Pichai about AI in search. Patel's (valid!) concern is that Google is using information from publishers, like The Verge, to inform AI-generated search results, which readers will click on instead of clicking links to, say, The Verge. Pichai argues that the AI-generated results actually lead to more clicks. One reading on this is that Google is basically taxing the content-creation industry in order to generate more search activity, whose upside they won't share. But it's also fair to ask if the content industry has been taxing searchers, by wrapping the one-sentence answer they're looking for in lots of extraneous text and ads. The middle ground is a healthy level of mutual exploitation from both sides.
• In Works in Progress, Samuel Hughes writes in praise of concrete. The decline of ornamentation in architecture is partly one more instance of the general tendency for economic growth to make stuff cheap and labor expensive—the specific "stuff" in question is the mass-produced kind; one-off productions are, economically, going to look more like services. As it turns out, while this story is true, it oversimplifies things—mass production of decorative objects is about as old as decorative objects themseles. It's a good case study in how complex and overdetermined economic changes are: some forms of architectural decoration did indeed get more expensive, some of them got cheaper, but also, the tastes of people making aesthetic decisions changed, a process that's harder to model.
• In this week's Capital Gains, we look at companies with a "secret weapon" like a unique technology they've adopted, and how this often helps explain a bit of their success, but doesn't explain all of it.
• And on this week's episode of The Riff, we discuss how reserve currencies work and the case against buying a house. Listen with Twitter/Substack/Spotify/Apple/YouTube.

Open Thread

• Drop in any links or comments of interest to Diff readers.
• Some industries are notoriously low-productivity because they're bottlenecked by labor. What are some examples of industries that are labor-intensive but have had surprising productivity gains over time?

Diff Jobs

Companies in the Diff network are actively looking for talent. See a sampling of current open roles below:

• A startup helping people build communities without losing ownership of their data is looking for a site reliability engineer with deep Kubernetes experience. (Remote, SF preferred)
• A well-funded startup is building a platform to identify compliance risks associated with both human- and AI-generated outputs. They are looking for a cloud infrastructure engineer to join their team of world-class researchers. (NYC)
• A top prop trading firm is looking for an intellectually curious, mathy generalist to work on projects spanning business strategy, technology, and markets. (NYC)
• A startup building the world’s most performant parallel-EVM is looking for a low level engineer with C++/Rust/CUDA experience who wants to deploy their GPU-EVM core to web3 projects. (Remote, EU preferred)
• A startup building a new financial market within a multi-trillion dollar asset class is looking for a junior ML engineer, especially someone interested in using LLMs to make unstructured data more tractable. (US, Remote)

Even if you don't see an exact match for your skills and interests right now, we're happy to talk early so we can let you know if a good opportunity comes up.

If you’re at a company that's looking for talent, we should talk! Diff Jobs works with companies across fintech, hard tech, consumer software, enterprise software, and other areas—any company where finding unusually effective people is a top priority.