Remains of the Day: Issue 10

Seeing Like an Algorithm

As a reminder, I’m Eugene Wei, a former product guy at companies like Amazon, Hulu, Flipboard, and Oculus, and I’ve kept a personal blog Remains of the Day since 2001, offering what I describe as “Bespoke observations, 80% fat-free”, though the perceived fat ratio will vary depending on your tastes in technology, media, and all the other random topics I cover.

---

I forgot to put an issue number on my last entry, which should have been my ninth. I’ll just mark this one as the 10th issue for some internal file naming consistency. This, of course, makes issue 09, which was unnumbered, a collector’s item. If you own one of those rare copies, put it in Mylar, it may be worth something someday.

Some updates since last time. First, I joined Julie Young as guests on Alex Danco’s blog to discuss the internet and gift culture. He has four entries on gift culture, all good. If you missed it, start with Part 1 and read forward. The post-Social Capital world has Alex blogging regularly, Chamath out there spinning up SPACs and railing against Facebook, it’s almost enough to consider Social Capital’s unraveling a, pun intended, gift.

Second, I realize I completely forgot to share a link for a 1-year subscription to the newsletter The Diff with an extra month free. My friend Byrne Hobart launched The Diff around the same time we locked down for this pandemic in the U.S., and his output has been both prodigious and consistently thought-provoking.

Like many of you, I’m subscribed to way too many newsletters now, they are piling up in my inbox. Every communications channel gets arbitraged to the point of suffocation, this was always going to happen. But the thing is, quality wins out in the long run, and Alex and Byrne’s writing are consistently novel and thoughtful, and that makes them the ones I open every time they land in my inbox.

In Byrne’s latest issue, he writes:

Countries accelerate their growth through “financial repression,” encouraging workers to consume less and save more, and recycling those savings into ever more investment. Sometimes, the policies behind this are fairly benign (a weak safety net and few worker protections will force workers to save much of their income to self-insure against unemployment), although some of them can be quite brutal (Stalin starved Russians to move resources away from agriculture and towards heavy industry). Startups also engage in a form of financial repression, by weighting pay towards equity instead of cash.

I’ve never thought of the startup equity pay model as a form of financial repression, but outside the tech bubble, almost no one really works on that compensation model. Very few employees outside of tech have so much of their earnings tied up in equity ownership, and in fact, most of them probably would suspect that type of compensation package to be a swindle.

In Hollywood, even some A-list, Oscar-winning actors and directors still think from paycheck to paycheck. If they are lucky, they might get a few profit participation points on one of their projects, but that’s an exception, not the rule. George Lucas is a sort of unicorn in the film business for having retained full ownership of his work and all its ancillary revenue streams (at least until he sold those off to Disney, an outcome which may or may not be one of those “be careful of what you wish for” cautionary tales).

Byrne also writes:

The End of History and the Last Man was and is a controversial book. Some of Fukuyama’s critics clearly didn’t read the whole thing, and some of them didn’t even make it all the way through the title. The term “The End of History” is pretty easy to interpret: all the big questions have been answered, capitalist democracy is the system everyone eventually adopts, and it’s hard to imagine a successor, so the big events are all in the past. “The Last Man” takes things in a darker direction: it’s a term from Nietzsche, referring to someone with no urge to create, no willingness to take risks, someone who has been spiritually poisoned by peace and comfort. The Last Man has lost all desire for recognition from peers, and is ruled by baser desires.

One of the more noticeable developments in the competition for talent in the tech industry the past decade or so has been the rising cost to startups of competing for talent. Google was the first of the FAANG companies I can remember to look at the copious cash on its balance sheet and ask, “Hey, instead of share buybacks, what if we just spent some of that cash directly on employee retention?”

As a competitor or even just a new startup, if you tried to poach a key Google employee, they’d take that offer to their bosses who’d come over the top with a retention bonus. Having lost a few potential hires to this tactic, I confess some grudging admiration for this type of ruthless tactic. HR departments always cite figures on how expensive it is to train up someone new to replace an existing employee operating at full efficiency. At some price, it’s just good sense to avoid that churn altogether.

The result, however, has been that the risk-reward curve for Silicon Valley has been flattened. More and more, it makes more sense to go to one of the massive, post-product-market fit tech giants, collect a really generous mid-sized salary and equity package, rather than go to a startup, where unless you’re really early, your ownership stake will be relatively tiny, and only a massive liquidity event will change your life (one could argue that is precisely the selection bias you want, aligning everyone from investor to founder to employees on shooting for the stars, but that’s a topic for another day). The low and medium risk outcomes in tech can be unbelievably generous if you land at the right company now.

The more pernicious side effect, though, is a rise in the number of middle managers who fit “The Last Man” description above. Resting and vesting at a Google or a Facebook, with no burning desire to create. “Spiritually poisoned by peace and comfort.”

In Part 1 of Alex’s series on gift culture, he writes:

Abundant environments may surprise you: even though they’re lacking in material scarcity or literal friction, there’s still plenty of work to do. It’s just a different kind of work: the work of dealing with complexity, clarity, curation, and especially synthesis. The effort and value being traded here lend themselves far more naturally to a gift culture economy, which is still very much an economy. It’s just not a transactional one.

The other obvious kind of work to do in an abundant environment, of course, is achieve and maintain positional scarcity. Status is clearly scarce, and in a gift culture like the free software community – or on Finance Twitter – the way you earn status is by putting in real effort, and then giving away the fruits of that effort.

I recently started making my way through The Lords of Strategy: The Secret Intellectual History of the New Corporate World by Walter Kiechel. It’s a history of four men who, in the 1960s, invented corporate strategy: Bruce Henderson, founder of Boston Consulting Group, Bill Bain, creator of Bain & Company, Fred Gluck, longtime Managing Director of McKinsey & Company, and HBS professor Michael Porter. Before they came along, a lot of business was just planning.

But, to tie this to Alex’s point on abundance, it feels like a lot of the strategy frameworks these four invented have become a bit outdated in the internet era, characterized as it is by abundance. Classical economics, with its supply and demand curves, to take one example of its equilibrium-driven prescriptions, works fine, for the most part, when you deal with physical scarcity.

But the internet often pushes us to the edges of classical economics, to the far ends of traditional formulas. What if supply is effectively infinite? What if a lot more of the economy is made non-rivalrous and non-excludable? What if more markets were global? What if everyone in the world had access to an information coordination mechanism called the internet?

A lot of traditional strategy frameworks aren’t serving entrepreneurs and business people well in a world where scarcity has increasingly moved from the physical to the digital sphere.

We’re still waiting for the new business strategy gurus of the information age, the new lords of strategy. There are hints of it when reading books like The Origin of Wealth or Brian Arthur’s work like Increasing Returns and Path Dependence in the Economy. But most the practical strategy advice for dealing with this new era come in bits and pieces in random blog posts from practitioners who’ve learned the hard way, on the front lines.

All of this is to say that I enjoy Alex and Byrne’s work in the ways bits and pieces of their thoughts converse with various things I’ve been turning over in my head.

---

Seeing Like an Algorithm

I’d be lying if I didn’t say I felt some sense of shame over not sustaining a high volume of output on this channel this year, but the volume of good work from others dampens that a bit. I long ago passed the point where I felt like we had a shortage of great work to read. Now the struggle is to find enough time to read everything piled up on my bedstand, in my Kindle, in my inbox. Every day the curation feels more ruthless.

Yet in particular areas, we still lack enough eyes or light. I find my posting frequency on my blog and here in my newsletter to be down, but I’m trying to compensate by looking at areas where I’d like to see more discussion.

You’ll get a good chuckle out of the fact that I’m going to follow that by pointing to my second of a 3-part series on TikTok. If there’s any company that seems over-analyzed at this point, it has to be that one.

I hope, though, that this latest piece Seeing Like an Algorithm helps people understand just how it is that TikTok’s FYP algorithm works as well as it does. I wrote in my last issue that TikTok’s algorithm is critical to its success, and that feels like a consensus view nowadays, but many have taken that to mean that the algorithm is some magic software model that the company couldn’t live without.

There’s another explanation, though, that I outline in this new piece. The algorithm is critical to TikTok’s success, but the algorithm itself is not some anomalous breakthrough in machine learning recommendation models. Instead, what’s magical about TikTok is that it’s entire design and operations provide that algorithm with a unique dataset on which to train itself.

Like GPT-3 or any number of vision AI models, TikTok’s algorithm is eerily good at prediction. But unlike those other models, TikTok’s model couldn’t turn to large, publicly available training datasets. The power of the TikTok flywheel is that its users use its app and tools to create the unique types of short videos that the app then trains its FYP algorithm on. TikTok has a closed loop of deep learning.

Anyhow, check out the piece for a deeper dive into just how that closed loop works, and what TikTok’s algorithm-friendly design has to teach other companies about “seeing like an algorithm,” a riff on James Scott’s Seeing Like a State.

I have one more piece on TikTok left to complete, but if you want a sneak peek at the thesis of each of my three pieces, I went on the A16z 16 minutes on the News podcast to outline them with my friend Sonal. I think we went over 16 minutes; as usual, that’s my fault.

I do think more podcasts during this long pandemic should start to involve a glass of wine. Maybe next time.

---

NY Film Festival 2020

I felt a particular twinge of sadness the past week and a half because it’s the first year in many that I haven’t been in Toronto for their fantastic film festival. Americans aren’t welcome in Canada right now, and TIFF’s virtual screenings are only available within Canadian borders.

NY Film Festival to the rescue. It is screening some of its selections virtually this year, and tickets are still left to some of those. If you purchase any of these, make sure to mark your calendar for what dates your films will be available to stream. I lost track of my purchase of Lover’s Rock from Steve McQueen and the viewing window closed before I’d started it. Luckily they added an encore screening. I’ve heard its excellent, and other films from McQueen’s Small Axe anthology are also screening at the fest.

Nomadland, from Chloe Zhao, fresh of winning the Golden Lion at Venice and the People’s Choice Award at Toronto, is sold out, but tickets are still available for the latest from Jia Zhangke.

I have a full slate of virtual screenings this week, and while it isn’t quite the same as gathering with friends in Toronto, I’m looking forward to lying down in the dark with my iPad and a bowl of popcorn.

---