Debunking blitzscaling

While Hoffman and Yeh’s book claims that companies like Google, Facebook, Microsoft, Apple, and Amazon are icons of the blitzscaling approach, this idea is plausible only with quite a bit of revisionist history. Each of these companies achieved profitability (or in Amazon’s case, positive cash flow) long before its IPO, and growth wasn’t driven by a blitzkrieg of spending to acquire customers below cost, but by breakthrough products and services, and by strategic business model innovations that were rooted in a future that the competition didn’t yet understand. These companies didn’t blitzscale; they scaled sustainably.

Google raised only $36 million before its IPO—an amount that earned Sidecar’s Sunil Paul the dismal third prize of going out of business. For that same level of investment, Google was already hugely profitable.

Facebook’s rise to dominance was far more capital-intensive than Google’s. The company raised $2.3 billion before its IPO, but it too was already profitable long before it went public; according to insiders, it ran close to breakeven from fairly early in its life.

Tim O’Reilly, The fundamental problem with blitzscaling

To my own mind the fundamental problem with blitzscaling is that it salts the fields. When an investor-funded company like Uber enters a market with a nonsensical business, it undermines the economics for companies in that same space. Cab drivers, for example, were able to make a real living before Uber.

Photo Morph Sorter

Given a randomly selected stack of photos, an algorithm could put them in order such that they appear to be morphing. What’s creative is that the morphing would be done without transforming any of the images.

Here’s a black car turning into a white cat.


The simplest algorithm would start with a set of random images and a similarity score between every pair of images, and a randomly chosen starting image. Then, given any previous image, the next would be the most similar unused one.

Choosing the starting image at random will lead to missed opportunities. In the example sequence above, if it had started with any image but the black car, the black car would have been left until last, and ended up appearing after the white cat. This would make no sense visually. It would create a breaking point in the sequence where everything went off the rails.

A different algorithm:

Start with a set of random images, a similarity score between every pair of images, and a randomly chosen starting image. Then, for any unused image, insert it into the sequence at the point where the sum of the distances between the inserted image and the adjacent ones is smallest.

However, at each step you might be planting problems for later. Another strategy would be to invent a metric for the chain as a whole, start to finish, and use a Markov Chain to find the optimal sequence.

(Idea for the metric:

  1. sum of all pairwise distances. This would capture places where a choice earlier forced a bad selection later
  2. largest jump between any pair of images).

Ok, I have to go to work. This has been an entry in an open notebook.

Podcasting Metrics

Podcasting needs metrics. There’s no reliable way to track listenership.

Advertisers need to know what they’re buying. The ad business is utterly reliant on hard data about listeners. They need to know how many listeners they’re reaching and what the listener demographics are. To prevent fraud by podcasters, advertisers also need provability.

On the content creation side, podcasters need to know what listeners like and don’t like. Youtube has a feature to show video creators when viewers drop off. Podcasting has nothing like that.

These are solvable problems. A simple-ish way to create metrics is to provide streaming audio rather than downloadable. Technically this is well-known territory. Mozilla has excellent documentation.

It’s simple in theory, but in practice there are hurdles.

  • This approach would exclude listeners who download their audio in advance. I doubt this is a big proportion.
  • Podcasting portals may serve up their own copies of podcast audio files, rather than redirecting to the original URL hosted by the podcaster. Streams can’t be cached.
  • Podcast listening tools may not support streaming MP3. How bad a problem this is depends on which streaming technology the podcaster is using.

Podcasters probably would lose listeners. How many listeners would they lose? They would probably be able to charge higher ad rates, and sell to more advertisers. Would that advantage in ad rates and sell-through outweigh the drop-off in listenership?

The new face of all-in-one subscription

Ben Werdmuller has subscription fatigue and thinks unbundling is untenable:

After more important expenses, there’s no way these are acceptable costs for most Americans.

ultimately, consumers will be paying huge monthly sums and subject to the bundling deals of whichever network they choose to be connected by, albeit with the ability to pay a la carte for additional subscriptions on top of our bundles. We’ll swap one set of gatekeepers with another set of gatekeepers.

I think he’s missing the simplest solution: only subscribing to one source. Netflix is insanely deep. A family could easily get by with nothing but paid Netflix and free Youtube.

If that’s the path the masses eventually take, we’ll have a situation like the desktop OS market, which has only three real competitors.

Unintended consequence: more delivery bots

The new bill to make app-based employers like Uber treat gig workers like employees has a workaround for employers: devices which can be remote-controlled by out of state operators.

Instead of UberEats having drivers deliver, they can use delivery robots. These bots are perceived as  autonomous, but in reality they are often remote controlled by workers in low-wage areas.

I’d give that 🤯🤯🤯 out of 🤯🤯🤯🤯

Trump has privately said that foreign spies can damage relations with their host countries and undermine his personal relationships with their leaders, the sources said. The President “believes we shouldn’t be doing that to each other,” one former Trump administration official told CNN.

In addition to his fear such foreign intelligence sources will damage his relationship with foreign leaders, Trump has expressed doubts about the credibility of the information they provide. Another former senior intelligence official told CNN that Trump “believes they’re people who are selling out their country.”

Even in public, Trump has looked down on these foreign assets, as they are known in the intelligence community. Responding to reports that the CIA recruited Kim Jong Un’s brother as a spy, Trump said he “wouldn’t let that happen under my auspices.”


As an act of seeding the open web, I created a blogroll of friends with blogs. Blogrolls are good for all blogs. Interconnectedness makes every blog more valuable.

It wasn’t easy to find living blogs. I had to web stalk a bunch of people. I went through my Feedly subscriptions to find real people. I also combed my Twitter follows.

The blogroll will only be useful on-site at, and not in RSS. For the sake of inclusion in RSS (just once), here is the list:

Link: “We Should Replace Facebook With Personal Websites”

(Vice) Personal websites and email can replace most of what people like about Facebook—namely the urge to post about their lives online.

Facebook isn’t really all that much better or more convenient than having your own website, or sending emails or chats. But for some reason, Facebook (and Instagram) are where we post now.

I don’t think personal websites are for the same purpose. Personal sites like this blog will never be the best place to distribute your thoughts to the biggest readership possible.

A personal site is a Moleskine, not a book.