Planned economies, Federated Learning, and Vimeo's turnaround

In which I write about Vimeo's remarkable turnaround story

Hey everyone,
Greetings from Washington, D.C!

This turned out to be one of my favorite issues to write. It checks all the mental checkboxes I have for this newsletter—the book of the week challenged me intellectually by coming at things from an unusual point of view, the long read of the week forced me to me explain something technically complex in plain English, and the business move of the week is an overlooked success story. As always, it is a pleasure to grace your inboxes with what catches my fancy every week.

Before we dig into the meat and potatoes of this issue, there are a couple of internet places I made an appearance on this week that I wanted to share:

  • Half a Thought by Abhinav Kejriwal: A wide-ranging interview that revolved around this newsletter and its creation process. Abhinav and I talked about my unhealthy caffeine habits, why building in public is overrated, and my note-taking systems (or lack thereof.)

  • Startup Salad by Nate Elliot: In his exploration of a new audio product, Nate writes about this newsletter and what I consider to be it’s backbone—the book of the week section—and how I approach preparing for.

I hope you check these out and if you do, please feel free to let me know your thoughts.

With those out of the way, let’s dig into this week’s issue of Snapshots in which I want to talk about:

  • The future of planned economies

  • The rise of federated learning

  • The (ongoing) biggest turnaround story of the decade

  • The dangers of SPACs, the wonderful world of Muji, and the making of a cello

Book of the week

Last week, we talked about the basis premise of The People’s Republic of Walmart—that giant corporations are planned economies and let local (at the department or store level) sub-optimals exist in favor of better global (at the company level) outcomes.

The second half of the book is equally thought-provoking. We get a brief look into Amazon’s planned economy with a focus on its warehouses the origins of the National Health Service in the UK, the factories of Foxconn, and the American military-industrial complex. All are examples of how planned economies are more optimal than markets when the good to be delivered is essential1. I also think that it does a decent—if not perfect job—of discussing the elephant room of how planned economies have historically been coupled with autocratic rule.

My takeaway from the book is that some level of planningng would be good for society—especially the more developed ones which have the luxury to do so. It is no surprise that pure capitalism can lead to sub-optimal outcomes. It’s tough to look at the fact that the US spends 17% of its GDP on healthcare2 and yet tens of millions of its citizens live on the teetering edge of medical bankruptcy and conclude otherwise. I’m no healthcare expert nor an economist, but I suspect that most people reading this agree with me.

The real question is to what extent industries should be planned. Here, I think the book falls short. In every single industry, there will be different levels of intervention needed and there’s no framework that the book provides to guide our decisions.

At the end of the day, what planned economies have against them is branding. It makes me think of the branding problem that cryptocurrency has—someone like my dad or older mentor think about it as a tool to buy illegal drugs or a nerd-in-the-basement niche interest. Similarly, as soon as you say the words planned economy and communism in a sentence together, political operatives and voters set off proverbial (and sometimes literal) alarm bells. McCarthyism3 may have been stamped out by Eisenhower’s hidden-hand4 and the Senate’s “cooling saucer,”5 but it tapped into something very real—a visceral opposition to anything that comes with sickle and hammer branding on an ideological basis, no matter what the actual idea.

I optimistic here. I suspect that the entrenched economies will change at some point and become more egalitarian but this will almost certainly be triggered by a crisis. Entropy usually reigns in our lives, but crises create those rare alignments between political will and political skill that large-scale reforms are born from.

Long read of the week

Federated Learning: Explained

While machine learning nerds have been talking about Federated Learning for a long time, I have seen an increasing number of folks outside that niche talk about it so I thought it would be useful to have give a quick background on what it means and its implications.

If you took an introduction to machine learning class, you’ll learn that the “learning” happens when you train a model and data6 on the same computing resource—think a large data server in the middle of nowhere. The trained model can then be sent to a device—think your phone—where it will do a pre-specified job. This job could be something like removing the red-eye from that Friday night party photo from 2019. Overtime, the use of this red-eye remover will generate data on your phone which helps improve the algorithm for you.

There are two problems here:

  1. The initial centralization of data to train the model creates significant privacy risks.

  2. The improvement of the algorithm on your phone doesn’t help others. It remains localized on your phone. An easy around this might be sending your data to the server so it can train on it. But you don’t want to send your data to a server—your data is especially vulnerable to hacking and other concerns during data transfers of any kind.

Federated learning focuses on the second problem.

Instead of sending your data, your phone computes the improvement to the algorithm using your data locally (on your device), and sends the marginal improvement as an update back to the server. This way, your improvement—along with improvements from every other phone using this algorithm—can be incorporated in the overall algorithm and sent as an update to every phone.

The upshot? Your data remains on your phone and everyone still get a better, continuously improving algorithm.

Business move of the week

Vimeo: One of the best turnaround stories of the decade

If you’re like me, you think of Vimeo as the place that indie films are hosted never to be watched by anyone. And for years, noting that Vimeo was a serious competitor to YouTube would elicit nothing more than a smug smirk from Google executives. Even as soon as 2018, Vimeo was seriously struggling.

But while YouTube has continued it’s march towards medium-length video dominance—TikTok/Snapchat straddles the Google-owned company from the short-form end and Netflix/Disney+/HBO straddles it from the longer form end—Vimeo has decided to play a different game altogether.

It is targeting small and medium-sized businesses which are looking to create video content. This includes everything from marketing material for various social media platforms to training material for employees getting onboarded. At the center of it all is Anjali Sud, the CEO of Vimeo since 2017, and the instrumental figure in this turnaround. Here is her take on the vision of this new Vimeo from an interview with Protocol7:

We want you to make content — and by the way, we're working on using our API and our data to actually tell you, "Here's a video you just made, we just optimized it for TikTok, Instagram, Twitter, LinkedIn and Facebook. Press this button, and each version will go out, and you can see what you want to do with it." We're going to automatically optimize the thumbnails and do X, Y and Z. The real thought here is that right now, the platforms are asking businesses to choose, and you shouldn't have to choose, and we don't want you to have to choose.

I think people under-appreciate how powerful this is. When discussing why they are not on some platform, they’ll typically say something along the lines of “it’s just not me.” But if they do try it, they’ll find that it’s very difficult to master even one medium, let alone multiple. That’s why corporations employ dozens of channel-specific managers. But small businesses don’t have that luxury. Vimeo helps them do exactly that. To use Shopify’s lingo, they are arming the rebels8.

There’s another story here—that of complexity convection. As I co-wrote with Nathan Baschez9 last year, companies evolve to solve ever more complex problems by continuously searching for the most profitable customers at the margins. In this case, Vimeo has moved from up-and-coming film artists—not exactly the most lucrative users—towards businesses which can actually pay for the differentiated services it provides.

Vimeo’s turnaround has been so dramatic that it is on it’s way to become an independent company (it’s currently owned by IAC) and reap the benefits of being an independent entity—most importantly, to operate without the shackles of keeping an eye on IAC’s interests. It is a perfect execution of the IAC strategy10.

I’d love to hear any thoughts here. My upshot here is that turnarounds are relatively rare in business and exceedingly so at this scale. We should celebrate them when they happen.

Odds and ends of the week

Two articles and a video this week:

💸 The Dangers of SPACs: Some writers are optimistic. Some are more pessimistic by nature. I try to be structural in my approach without a bent towards good or bad. So with that in mind, here’s a rule of thumb: when a new tool comes along and money is involved, people will bend around the technicalities to make money from others in a way that is different from the original purpose of the tool. This NYT article lays out how that might play out for SPACs by way of misaligned incentives. Regardless of where you stand on SPACs (I think they are largely positive), you should understand the opposite narrative.

🇯🇵 The Commercial Zen of Muji: When I lived in Manila, there was a set of inner-connected malls across the street from our apartment. And on the days with torrential downpours (i.e. for entire months of the year), I would use an intricate set of underground crosswalks to enter this web of capitalist dreams. And my first stop would be the Muji store. I liked it because of a few things: they had everyday items with minimalist designs, the stationery was top-notch, and it was just beyond the causal purchase price point—buying something from there was an experience that I would including saving up or cajoling my parents to buy it for me. It was an exercise in practicing either my delayed gratification or my negotiation skills. Anyways, this article goes into the history of the company and how that past is inexplicably linked to its future. I enjoyed reading about it.

🎻 The Making of a Cello: Get lost in the world of Italian craftsmenship with a lucid soundtrack. I could watch these “Making of” videos for hours.


Of course, the meaning of what is essential changes depending on who you ask.


See this list from World Health Organization Global Health Expenditure database.


Colloquially, McCarthyism was the phenomena of charging someone without evidence as an ideologue of some sort—typically as a Communist—without any evidence to smear them. Originated during the early 1950s by the branding of US officials as Communist by then Wisconsin Senator Joseph McCarthy.


See the issue on The Hidden-Hand Presidency.


See The Master of the Senate, Chapter 23: Tail-Gunner Joe by Robert Caro


A simple example: If I give you a,b,c and ask you to predict d, then a,b,c are the data and the logic you use to predict d is the model.


See my post on Shopify’s corollary to this—financing the rebels.


See Complexity Convection on Divinations


See this press release by IAC from December, 2020.