github twitter linkedin instagram email
Embrace, Extend, Extinguish
Feb 17, 2020
7 minutes read

GitHub just released their gh CLI, with initial support for issues and pull requests. While it’s complimentary to git, gh squashes a bunch of otherwise confusing checkout and pull commands and makes the interface much cleaner.

It’s a slick experience that I’m sure will continue to see growth, and at some point gh commands will start making an appearance on GitHub docs and eventually on the main site, for example in the “clone or download” tab on a repo, the default could be gh clone org/repo instead of using git over SSH.

As I was reading the HN commentary, I stumbled upon a great question, “why doesn’t GitHub just use files in a git repo to add issues, pull requests, etc.?” I shared these thoughts in our group chat with the joke “embrace, extend, extinguish” added on, a reference to the Justice Department’s description of mid 90’s Microsoft’s strategy (many examples). Given GitHub’s ownership, I was poking fun at what the future might hold, as several others on the original thread did.

To be clear, I’m not calling the death of git, nor am I particularly mad that people are building more user friendly tooling on top of it, but it did get me thinking about building good products with defensible business models. This post aims to condense those thoughts, as well as point out a few other areas this seems to be happening in.

Embrace + extend = extinguish?

As discussed in week 3 of my PM course, companies with good business models usually have certain advantages: the founders are best in class, their products are 10x better (or cheaper), they have a distribution network that lowers their customer acquisition costs, or they’ve built up moats that keep competition at bay. Embrace, extend, extinguish is one of the easiest moat building techniques companies to implement. Here’s how I’d go about building a company with a good moat.

First off, you need a protocol that is mature enough to have significant usage, but also crufty enough that features can’t be easily added to it. It also needs to provide backwards compatibility so legacy users aren’t left behind.

Some examples:

  • git (version control)
  • smtp (email)
  • sms (text messaging)
  • sftp (file transfer)
  • rss (content aggregation)

Next, you need a layer of abstraction to cover the implementation, that you can also run your proprietary API alongside. Email and SMS clients have done this forever, though they typically don’t deviate too heavily from what’s provided in the standard. I say typically, because the truly great ones do.

Once your product has sufficiently abstracted the underlying protocol, people slowly forget how the original worked, where the warts are, etc. and become increasingly reliant on the higher level of abstraction and backwards incompatible new functionality.

In the case of gh, once they have commands that take the 80% of git actions that involve multiple commands that nobody remembers, and wraps them in more relatable GitHub termed actions, people will forget the ancient ways.

This lays the groundwork for the final step: replacing the legacy protocol with your new one, and eventually removing it altogether. Once your tool has subsumed all that functionality, there’s no reason to keep the old one around. You don’t even need to consciously extinguish anything: your users effectively did that work for you.

More than just Microsoft

Examples of this are happening all over right now, and it’s not just large tech companies (though, most of the examples of it working only happen at certain scale).

iMessage: the 🐐 GOAT 🐐

As we reach the top of the S-curve of cell phone adoption, Apple has switched to accessories (e.g. AirPods) as well as services. Apple TV is an obvious play as is Apple Arcade, but so is the less obvious “Sign in with Apple.” The fact that it is required to be included when an app includes a NASCAR login screen is a move I’m surprised hasn’t been challenged. Once you sign in with Apple on one website or app, it becomes basically impossible to use the same account on a non-Apple device, because no Android app is even able to support signing in with Apple. These services all work to keep users buying Apple devices, as they only work on Apple hardware, and each add a new layer to the moat around that core business. And of all of them, iMessage is the best example, in part due to strong social pressure and network effects.

Google has tried to fend this off with RCS (its own extension of SMS, sort of backed by the telecom industry), but while that brings certain Android devices up to feature parity (“darker blue bubbles”, read receipts, typing indicators, etc.), it doesn’t solve the fact that non-iMessage folks will always be green bubbles in iMessage.

FaceTime is another great example. Remember in 2010 when Steve Jobs (pour one out) promised it would become an industry standard?

Now, FaceTime is based on a lot of open standards — H.264 video, AAC audio, and a bunch of alphabet soup acronyms — and we’re going to take it all the way. We’re going to the standards bodies starting tomorrow, and we’re going to make FaceTime an open industry standard.

Well, here we are a decade later, and I can’t use FaceTime on the machine I’m writing this blog post on. And I’m not surprised: the deep integration and walled garden of the Apple ecosystem keep selling hardware, and at the end of the day, as much as Apple wants to be a software (or now, a services) company, it’s still in the phone business.

Spotify: the home of podcasting

Podcasts are typically distributed via RSS feed to podcast clients, which are able to aggregate supply, provide the listener with useful functionality beyond playing, and also handle ad injection, etc.

Spotify, providing a client as well as an owner of several podcast groups, is able to aggregate both supply and demand, while adding additional functionality (e.g. better demographic information for ad targeting).

The worry here is that Spotify gains control of the ecosystem and forks from that standard (or the standard is otherwise rendered useless), and lots the hallmark diversity of the podcast ecosystem is lost.

Superhuman: coining the term “luxury software”

Superhuman is the iMessage to Gmail’s Android (though it actually uses Gmail’s APIs under the hood!). While Superhuman currently lags Gmail (and many email clients) in features (for example, formatting is very difficult/non-existent: try monospace fonts!), but they’ve critically added the blue bubble to Gmail’s green: Sent from Superhuman below the fold.

Users on Superhuman also have a “user since 201X” in their contact card, and it’s only a matter of time before they start adding Superhuman-to-Superhuman only functionality (e.g. reaction emoji to email threads or realtime chat). In effect, Superhuman becomes the online equivalent of The Battery, and those not present on the tool get left out of the real world dealflow.

That’s a powerful enough incentive to keep paying $30/month, especially since Bloomberg charges just over $2000/month for the finance equivalent (chat, meaning access to the network, being a huge component of the worth). I won’t write any more about Bloomberg, but if you’re interested I highly recommend “You Can’t Kill the Bloomberg Terminal”.

Replacing moats full of money

The past several years have seen startups use cash as the moat: e-scooters, food delivery, mattresses (and almost every other DTC vertical). All of these are commodity goods, with zero defensibility: the motto was, “the best defense is a pension funded offense.”

Given the high profile dumpster fires that have come out of that and lackluster IPO performance of the ones that made it that far, I hope we’ll see more startups picking defensible positions and having moats grounded in high quality experiences and engineering choices that result in faster, better, or cheaper products.

As always, I’m interested to hear your thoughts, and if you know of any other good examples of this, especially among early stage startups, definitely reach out and let me know!

Back to posts