115: Cloudflare (part 2), Microsoft Gets Nuance ($16bn), Does Spotify Even Like Joe Rogan?, SpaceX Starlink, Google's ML Codec, VR, Pfizer Factory, and Pufferfish Skeletons
"you’d be downright irresponsible not to"
Beware those who claim your argument is ridiculous.
Seek out those who claim it is untrue.
☯️ 📚 I’m sure there exists many “book pairs” where 2 + 2 = 5.
The one that immediately comes to mind:
'The Outsiders' by Will Thorndike
'The Halo Effect' by Phil Rosenzweig
Each is quite good on its own, with some flaws and blind spots. But together, they balance each other out and create an oeuvre on a different level of Yin & Yang.
What about you? Any good book pairs that come to mind (or trios / quartets / quintets / sextets(!!!))?
Hit reply to the email, or write in the comments on the site to let me know. Thx
🐜 I was thinking about skeletons vs exoskeletons… While I understand why exoskeletons don’t make sense at our scale and have plenty of downsides that would’ve limited our evolutionary flexibility greatly, on some primal level, they make a lot of sense and are really cool.
The world is a hostile place.
Having your squishy, vulnerable bits on the inside rather than on the outside has lots of advantages.
Good thing we’re smart enough as a species to invent all kinds of ways to tailor the world to our needs and protect ourselves, because we’re just a bunch of meat stuck to sticks
Modern human history is basically us child-proofing the world by removing dangerous things, making everything flat and even, controlling the temperature, wearing protective gear, etc.
💚 🥃 I figure that the price of a couple coffees or one alcoholic drink isn't a bad trade for 12 emails per month that take hours to research and write.
The entertainment has to be worth something, but for those that care most about the bottom line: if you make just one good investment decision per year because of something you learn here (or avoid one bad decision — don’t forget preventing negatives!), it'll pay for multiple years of subscriptions (or multiple lifetimes, for the Capital Allocators reading this).
As Bezos would say of Prime, you’d be downright irresponsible not to have a subscription1:
Investing & Business
More Cloudflare (…a lot more)
Can I have two editions in a row with lots of Cloudflare stuff..?
Hmm, I guess so!
Here’s a grab-bag of highlights from their recent analyst/investor day (Koyfin has the transcript):
Long term, we want our Workers edge platform to be as rich and powerful as any of the centralized cloud providers, but maintaining our differentiated advantages of production readiness baked in, taking advantage of Cloudflare's network and all of the products we built on top of the platform.
That’s a very ambitious goal. Looks like there’s no plan to stop expanding the functionality of their stack any time soon.
In 5 years, if you're connected to the Internet and you're not using Cloudflare One, you're doing it on your own.
We're totally happy to coexist with boxes that customers want to keep or still have. We're going to win this war of attrition over time because our products are better, our roadmap is better, our ability to execute is better, and we're really, really patient.
This is something I see over and over again with this company: Evidence of long-term thinking.
They won’t even try to define the TAM of their Workers platform yet (years after first unveiling it) because it’s still emerging and they don’t want to constrain things too early, they offer products for free and take their time monetizing them, they constantly talk about the the grand vision of making the internet better rather than focus on just specific products or industries, etc.
Most other companies try to convince you that they’re in a really good place right now, while Cloudflare seems to want to show that they’re going to a really good place in the future.
Last March, we started giving all of our products away for free and the Teams there in the zero trust area. We learned a huge amount, and we actually gave them away for free for 6 months of last year. What have we learned from that? We learned that our onboarding was too difficult. [...]
We ended up actually converting 75% of our enterprise customers to paying from free. [...]
we launched a suite of products, which we call Teams Standard, which gives 50 seats free forever for small businesses. And that came out of the learning from last year also in 2020, and that's now giving us free customers that are moving into our pay-as-you-go area.
More below on how all this free stuff gives them a data and product dev feedback/testing advantages, but it’s also doing something else that they probably won’t brag about:
It removes oxygen from competitors.
If you can use a bunch of Cloudflare stuff for free, it makes it a lot harder for smaller scale competitors to bootstrap themselves.
That, plus some of these small free accounts will grow into big businesses tomorrow, and it only takes a few big wins to pay for a lot of free stuff.
At Cloudflare, we push code every day. I mean, last year, 2020, what was it? I think, we were really something like 500 new capabilities to customers around the globe. [...]
At Cloudflare, we believe in getting into market early and learning and growing to be that best-to-breed market-leading solution. [...]
we don't have data silos. All of our data is available to all of our products. We are built on the same platform, which means that we inherit the speed and security of that platform.
Here I hear echos of Elastic, Crowdstrike, and Snowflake (for different aspects each).
data localization and data sovereignty. More and more of our customers are confronting patchwork of regulations in different countries [...] the fact that our network is everywhere and our storage products are everywhere, makes it really straightforward to solve those sort of micro compliance obligations in a really seamless way.
Helping customers deal with complex & cumbersome regulations is a tried & true way to provide value.
Another growth driver is our investment in enterprise sales. [...]
We've added over 650 new large customers since 2017, and this large customer cohort has increased a compound annual growth rate of 73%. The revenue contribution from these large customers is growing even more quickly with a compound annual growth rate of 90% since 2017. As the average spend of our large customers continues to increase, so does the mix of revenue from large customers, which represented 46% of revenue in 2020 and 49% of revenue in the fourth quarter of last year.
90% revenue CAGR since 2017 for large customers. That’s kind of bonkers.
Our inverted go-to-market model allows us to invest behind the demand and success we are seeing. We don't open an office in a new market until we see enthusiastic developers and revenue from paying customers. This approach removes an element of risk associated with international expansion and allows new hires in these markets to ramp quickly. As we gain traction in new markets, we've been able to move upmarket, improve our platform with larger and larger customers.
They mentioned this elsewhere too, but it’s really nice to be “pulled” into markets and verticals rather than have to “push” into them. This reminds me of Everbridge’s CEM. They launched in some limited geographies/verticals, and big customers elsewhere are basically begging them to launch in their areas.
At the IPO, we shared that 70% of contracted customers used 4 or more products on average. Today, 88% of contracted customers used 4 or more products. And furthermore, today, 79% of contracted customers used 5 or more products.
As customers utilize more and more of the platform, it drives our strong dollar-based net retention, which ended Q4 at 119% and remain between 115% and 119% in 2020.
This land-and-expand is tried and true. The first module is the one with the highest CAC, but once you have a relationship with the customers, any other modules you sell them are usually much more profitable (though the devil’s in the details, some services are just more expensive to deliver).
The amount of global traffic on our network has increased by 72% in 2020 alone.
since the IPO, we've added roughly 1.4 million free and paying customers, representing an increase of 64%. Despite these significant increases, we have been able to maintain an average network CapEx of 12% of revenue in 2020 and have consistently maintained high gross margins, averaging 78% since the IPO.
I think this is the beauty of being able to colocate inside ISPs in so many places. A lot less expensive than building your own datacenter, and a great moat since it’s not something you can just buy into, you have to be invited in (and Cloudflare’s early products like CDN and DDoS protection were win-win for ISPs, so they loved having them in the perimeter — especially in countries with data-caps where making the internet better and faster leads to higher data consumption, and so higher revenues).
we don't totally know what the next 10 years are going to bring. But we've built an engine that listens to customers, that sees everything which is going on, on the Internet. That sees opportunities and then can rapidly iterate on those in order to deliver those to our customers. So I'd take the over on an over/under bet on whether in 2021, we released more features than we did in 2020. And I take the over on an over/under bet on whether 2022 is going to be even more.
Talking about some of their consumer products, like the 220.127.116.11 Mobile App (I use it — at home on wifi I have it disabled because I trust my ISP, but when connected to cell towers I have it auto-activate and get me on Cloudflare’s VPN).
I'm really proud of is the fact that even though we're an enterprise B2B company, we’ve built products that consumers love. And when they don't love something or when you slow down the Internet or when you break something, they let you know immediately. And that creates this incredible feedback cycle that allows us to then build much richer products for our enterprise customers.
He’s right. I’ve used the “send feedback” feature like 15 times (including on the desktop version of WARP), and I usually get a email back from someone at Cloudflare. They’re very responsive and listen to feedback.
we really have tried to follow what we think of as the most natural path for SaaS companies, which is that you start with the low end of the market, that's completely underserved, and then you move up in the market over time and that we're following the footsteps of great companies like Salesforce. If you signed up the Salesforce in the beginning, it was CRM for small businesses. They send you a copy of Zig Ziglar's 6 secrets to selling. And again, but today, obviously, they have market penetration across the -- all of the large companies and have completely dominated that space.
Is that true? Anyone here got that book after signing up for Salesforce back in the day?
Friend-of-the-show Muji even got a question answered in the Q&A section, with a shoutout from Cloudflare CEO Matthew Prince (trying to remember how many “H”s at the investor day… nice):
Q: The next question is from Muji at the HHHYPERGROWTH blog. Can you expand on what Cloudflare is seeing for how customers are using Durable Objects on Workers?
A: Definitely, my favorite Twitter follow. And if you're a Cloudflare investor, [and] you're not following HHHYPERGROWTH, I -- it's H with -- it's hyper with, I think, 4 Hs (sic) [ 3 Hs ], so definitely check it out [...] it turns out that in edge computing, the computing part is the easy part. The hard part is the data storage part. [...]
the reason that data storage is hard in a distributed fashion, you can think about if you were selling like concert tickets. It turns out that you can't sell the same seat to 2 different people [...] you have to keep accurate control of transactions that's a very difficult computer science program, which is sort of framed by what's known as the cap theorem, which says that you can't have consistency, availability and partition tolerance. You can have 2 of those 3 things. [...]
we had a really good insight from the team behind it, which was that, for the most part, data follows people and people are in physical locations around the world. And so if instead of saying, I'm going to put all of my data centrally in 1 database, if you said, I'm going to put data about various users near where those users are. And then if the users move around, then the data will migrate to wherever they move to.
Reading all this reminded me of Patrick’s interview with Cloudflare co-founder Michelle Zatlyn. If you haven’t heard it, I recommend it, it gives a good idea of their leadership and culture DNA:
Michelle Zatlyn - Protecting the Internet (podcast, 58 mins)
I love what she says here about the creation of their ‘skunkworks’ R&D lab:
early on, what would happen is we had one R&D organization, one product engineering organization.
And what happened is, when the priorities came, all of the things that were over the horizon got deprioritized for things that were known quantities now, like, "This is part of our core service, we know there's revenue attached to it, we should be doing this now." And that's great for now, but you're mortgaging your future if you keep doing that.
Spotify Censored 42 Joe Rogan Episodes (so far…)
Spotify has now deleted more than 40 episodes of “The Joe Rogan Experience” from its archives, after the audio streaming giant recently — and quietly — removed two more episodes from its service. (Source)
Rogan has to be regretting his decision, especially when he didn’t need a middleman between him and his audience and could probably have made similar money, if not more, with fewer headaches, by going direct with subscriptions.
People say stupid things and have idiotic beliefs all the time. There’s plenty of stupid or untrue claims on Twitter and Facebook every day. There’s endless crazy on Youtube, and hosted on various blogging platforms like Wordpress and Blogger.
Spotify itself is full of gansta rap albums with some pretty wild lyrics.
Do they really want to be the thought police? The 21st century Nancy Reagan?
I get censoring racist hate speech, calls to violence, etc. But I don’t think this is quite what this is.
One of the latest episodes deleted, as first reported by Digital Music News on Tuesday, was episode No. 411 with Bulletproof Coffee founder Dave Asprey. The podcast episode initially aired in November 2013 [...]
It’s unclear why the episodes have been purged, but as DMN pointed out, “One explanation for the removals is that Asprey frequently backs controversial anti-aging and scientific theories, including claims that Bulletproof Coffee is extremely healthy while other coffee brands are not.”
So it looks like they deleted it because he made dubious claims about his coffee and diet?
I can only guess that Rogan pushed really hard in his contract not to be under editorial control, yet this is still happening.
There’s got to be tension there, and other podcast stars have to be looking at this + how Rogan seems to have lost a chunk of audience and isn’t nearly as in the zeitgeist anymore since he went exclusive (and off Youtube, which is likely a big part of this — it's almost like Rogan was banned from social media, except it's self-inflicted)…
I bet it’ll be a lot harder for Spotify to get big names to go exclusive in the future, and I wouldn’t be surprised if Rogan didn’t renew his contract when it runs out.
Everybody talks about the money he’s making as if that cures all, but he was already very rich and making millions. People like him don’t get to where they are, do multi-hour interviews every day, because they are motivated primarily by money. I bet his audience/reach and freedom of speech are higher priorities in the stack for him, so that has got to hurt.
h/t Jerry Capital
Microsoft + Nuance
definitive agreement under which Microsoft will acquire Nuance for $56.00 per share, implying a 23% premium to the closing price of Nuance on Friday, April 9, in an all-cash transaction valued at $19.7 billion, inclusive of Nuance's net debt [...]
Nuance is a trusted cloud and AI software leader representing decades of accumulated healthcare and enterprise AI experience. Mark Benjamin will remain CEO of Nuance [...]
Microsoft's acquisition of Nuance builds upon the successful existing partnership between the companies that was announced in 2019. [...] The acquisition will double Microsoft's total addressable market (TAM) in the healthcare provider space, bringing the company's TAM in healthcare to nearly $500 billion. (Source)
I don’t know that much about Nuance, but it certainly seems to fit well in Microsoft’s portfolio, and the most interesting part of the business seems to be the healthcare product. Since Microsoft has been collaborating on it, they know very well its potential and value, so the risk of bad surprises is probably lower than with the average M&A deal.
If you want to learn more, Ari Lazar wrote a piece about it recently:
[NUAN:] Screens bad (spinoffs, cashflow>NI, most debt convertible ~1% rate, margins growing), A+ CEO, >50% market share in DR dictation software, cloud transition for core products, new AI powered product DAX is game changing offering 10x ROI / >10x upcharge + solving DR burnout [...]
It is clear from the whitepaper that DAX is a revolutionary product. It increases physician efficiency, improves quality of care, and decreases physician burnout. For these reasons, DAX has the potential to meaningfully increase Nuance’s revenue. [...]
From everything that I have read, DAX is the leading product on the market for fully automated clinical note creation. Additionally, Nuance has recently acquired the distant second place contender, Saykara.
h/t Elliot Turner
Science & Technology
Video Tour of Pfizer Vaccine Factory in Michigan
Cool stuff. I was particularly impressed by how they expand the factory with modular rooms that were prefabricated off-site in Texas and then shipped to Michigan (just rolled into place).
I just learned about pufferfish skeletons…
VR Headsets in the Wild
it took 41 months for Steam to reach 1 million monthly-connected headsets but just 11 months to add nearly another 2 million for a total of 2.92 million monthly-connected headsets. Year-over-year, the number of headsets is up 110%. (Source)
Setting up Starlink (SpaceX) Satellite Internet
SpaceX’s constellation of low-Earth orbit satellites is currently in beta. Here’s someone’s experience with it:
There’s more discussion of it and personal anecdotes with the service in this thread:
I'm using Starlink right now. AMA.
I'm in East Idaho.
My speeds are inconsistent, and interestingly they start slow (around 60 Mbps) but after a couple seconds they'll get to 150-200 Mbps (which is awesome for downloads). Latency is consistently in the low 30ms. I get some downtime every day, so it really is a "beta" like they say. I have a backup WISP.
Setup was literally take dish out of the box, insert into tripod (included), plug in cables, connect to the wireless routers SSID and activate with the starlink app. After that I put the included router into storage and plugged in my Protectli running CentOS. Everything works great. My only complaint is the CGNAT, but given the difficulty associated with procuring IPv4 addresses, it's understandable.
Correction: Wow, for some reason, my brain originally read the location in the original post as Johannesburg (South Africa), but it’s actually Jonesburg, Missouri.
I can’t correct that in the email that went out, but I’m correcting it on the site. Sorry about that.
Lyra: A New Very Low-Bitrate Codec for Speech Compression (using Machine Learning)
Cool project by Google (which they are open-sourcing), which will be especially useful in large parts of the world where bandwidth is scarce and/or expensive:
we’ve developed a new model capable of reconstructing speech using minimal amounts of data. Lyra harnesses the power of these new natural-sounding generative models to maintain the low bitrate of parametric codecs while achieving high quality, on par with state-of-the-art waveform codecs used in most streaming and communication platforms today. [...]
Lyra is currently designed to operate at 3kbps and listening tests show that Lyra outperforms any other codec at that bitrate and is compared favorably to Opus at 8kbps, thus achieving more than a 60% reduction in bandwidth.
To really get an idea of how good it is, head over to this page and listen to the audio samples that compare Lyra at 3kbps to Opus at 6kbps down the page. It’s impressive (though I dislike that the Lyra samples are louder — this isn’t a fair comparison). I wish they had also compared it to ‘Codec 2’, an open-source low-bitrate codec that seems to do quite well.
Maybe someday Google will have technology so advanced that they’ll create a ML-optimized codec that makes earnings conference calls sound decent.
The Arts & History
‘La Soufrière volcano on the Caribbean island of St. Vincent erupted on Friday after decades of inactivity’
Just a Cool Portrait
Why pay when you can read for free? Well, think of me as a street performer. Don’t get me wrong, I enjoy juggling, but I have bills to pay, so if nobody puts anything in the hat, I’ll just have to move on…