118: Cloudflare (again?!), Bezos' Bakery, Coinbase Pro, Crowdstrike, Nvidia 2009, Snowflake for Dummies, Video Games, Canadian Real Estate, and Lord of the Rings

"here’s a crocodile"

There have never been so many opportunities to be told what to think by people who don’t think.

—Stoic Emperor

📝 I have lots of note files, used for different things.

I have notes for this newsletter in Notion, my investing journal in a Pages doc (equivalent to a Word document), and some ‘to do’ and ‘to read’ stuff in Apple Notes. I also have shared calendars and Microsoft To Do lists with my wife for shopping lists and ideas of TV series/films to watch…

I find that segregating notes by function among different apps helps keep things organized and mentally separate.

If I was to put everything in the same app, it could work, but it would feel a bit more overwhelming, at least to me (of course, your mileage may vary).

A recent change that I quite like results from thinking about my note system not by topic/task/job-to-be-done, but rather by “latency” and “level of organization/structure”.

I created a new note file called “working memory” (I got the name from Cal Newport).

It’s basically a plaint text file where I can dump any unstructured piece of info that I want to remember or may need again soon, but can’t classify in longer-term storage right away. It’s always open on my desktop on the side of the screen, and I can quickly access it on my phone too (I picked what loaded quickest, so it’s Apple Notes — Notion is way too slow for this).

The latency lens comes from computer architecture. If my investing journal is more like writing to HDD/SSD because I plan to keep referring to this data for years, and the Notion files are more like RAM (mostly days to months), this new thing is more like a CPU’s L1 cache (usually within the same day).

It’s designed to hold fewer things, but be very fast to access (read or write), and when I have the time to go through it, it’s designed to be flushed, either to the void, or to longer-term storage elsewhere.

How much thought have you given to your note-taking/storage system? If you work a “knowledge” job, seems like it’s important, and improving it — even a little — has high leverage.

📚 Follow-up on the book pairs that I mentioned in edition #115:

Reader Andrew Felmdan:

A pair I used to love recommending is:

- The Big Short

- Too Big to Fail

BS gives great insight into the micro of the GFC while TBTF provides the pairing macro narrative (not macro economic, but the large banks, regulatory agency responses)

That’s a good pair, I like this. Good idea to combine macro and micro this way. I’m sure there are other interesting complementary pairs that can be found through this lens.

f.ex. A pair that I quite like:

• With the old Breed by E.B. Sledge (memoir of a U.S. marine in the pacific theater during WW2)

• Strong Men Armed by Robert Leckie (high-level, large-scale overview of the USMC in the Pacific theater during WW2)

🐊 I didn’t know what to use to illustrate this one, so here’s a crocodile (or is that an alligator? *squints at the screen*).

Follow up on Asch Negatives from edition #116:

Friend-of-the-show and supporter (💚 🥃) MICapital wrote about “coming from a family of Asch negative people” and how it’s not all good times. “in general, I think it’s a bad thing. Or at least bad for the person who is Asch Negative”, he wrote.

From Talking to Strangers by Malcolm Gladwell:

“Human beings are trusting engines. We are evolved to give people the benefit of the doubt. It’s the right move 99% of the time… If you are a paranoid person, your life is a nightmare.

“If you’re always fighting the crowd, it can be tough. IMO.”

Very good points, very interesting.

My reply to him (cleaned up and expanded a bit):

On Asch +/- : I’m guessing it’s probably like a spectrum, or a dial.

The extremes are probably where there’s most problems.

On one side, you conform so much that you don’t find or do what you actually love, or may even repress your personality so much that you just can’t flourish in life and are easily abused/taken advantage of by others. Most people are probably closer to this.

But on the other side, there’s probably people like you describe, who are the mirror image of the first and it leads to similar — if inverse — problems.

For these people, working on more trust is probably helpful to bring them where I think the sweet spot is, which is closer to the middle, where you can follow and trust when it makes sense, but still have enough ease at going against the grain when you see it’s necessary, and won’t just freeze up or be too timid about it.

If we had infinite mental capacity, the optimal strategy in life would be to judge everything on a case-by-case basis (which doesn’t mean not taking into account past experience, just not jumping to conclusions or using crude heuristics but thinking it fully through).

That’s not really possible, but I think it’s important to retain that capacity to switch back to that mode without friction when something trips our BS detector, and then be ready to go in whatever direction our best judgement points to rather than succumb to peer pressure or reflexive contrarianism.

🌬 I was playing with my kids outside, blowing bubbles in soapy water. They were chasing after them and bursting them with pointy sticks.

I had the shower-thought (🛀) that bubbles are basically temporary containers for human breath.

I promise this isn’t some kind of financial market metaphor.

💚 🥃 Please help keep the lights on for this project and consider supporting it with a few bucks:

---

Investing & Business

Cloudflare Back on the Menu (Spécial du Jour)

Some of you seem to have liked the multi-edition Cloudflare coverage recently.

Byrne Hobart picked up the baton and wrote a great free post on the company. Some highlights below, but you should read the whole thing:

Cloudflare is, in part, in the business of making the Internet work the way a techno-optimist in 1996 might have assumed it would.

That’s a good line. (Byrne know how write good)

The idea is a bit like with software. You write your first version in 1998, and then you keep adding features and fixing bugs and updating the UI, and over time you’d think that you’re building something better because you keep iterating and polishing it, but there comes a point where you can’t just keep building on the old foundations, they’re just not what you need 10 or 20 years later. (this is how you end up with WindowsXP stuff under the hood of your operating system 20 years later)

So you need to de a serious code refactoring, or maybe even start from scratch…. Sometimes you can keep the old and new versions available in parallel and sell both during a transition period, as the new version reaches feature-parity with the old one and users — who have complex workflows built around your stuff — have the time to get used to the new way of doing things.

This is common with companies that used to sell software in boxes and on shiny disks with a perpetual license model when they transition to a software-as-a-service offering (SAAAAAAASS).

Things are similar with internet infrastructure, both the software and hardware.

MacGyverin’ has limits.

The original design keeps being improved and added to for a long time, but after a while, some problems just can’t be duct taped or painted over anymore, and you need serious surgery to make things work well or to allow new cool stuff to be built.

This is a bit what Cloudflare does with the internet. For example, the net wasn’t built with security in mind, so Cloudflare created their own network that is designed in a secure way and they allow users (paying and free) to send and get their precious bits (0101010101) through it instead of the old dirty, rusty internet pipes dating to the roman era (these things are full of lead).

Similarly for speed or data localization/multi-national compliance or load balancing or latency requirements or whatever.

Anyway, I digress…

The Internet has been evolving for decades, and Cloudflare can look at it holistically and apply some intelligent design to route around inefficiencies and design problems.

So yeah, what I said above in a lot of words, but compressed into a sentence. Thanks a lot for making me look wordy, Byrne.

The stylized facts that drive Cloudflare's value are that fake traffic is far more scalable than real traffic, which encourages people to launch attacks on sites; and that anything scalable has a statistical signature that makes it detectable, but only to someone who operates at scale.

Again, I repeat myself, but Byrne is really good at compressing complex ideas well.

This is what Cloudflare’s CEO Matthew Prince meant when he said: “the entire Cloudflare network effectively is one of the world's largest sensor networks …  our largest customers are able to be protected before anyone even is aware that a new attack is in the wild”.

They can recognize large-scale distributed attacks because they themselves are larger-scale and more distributed than any attacker. They have eyes and ears everywhere.

Cloudflare built an uptime factory, then workers became an uptime factory factory, and with the Bezos rule they've codified the production of such things: an uptime factory factory factory. They are no doubt adding new layers of recursion even now.

On Glassdoor, a frequent complaint about the company is that it doesn't pay especially well. Meanwhile, the company says "last year, we had 200,000 people apply to work at Cloudflare. We accepted less than 0.5% of that." Building essential infrastructure for the Internet is an interesting project; plenty of web pioneers did it for free, or for the price of a grad school stipend, so if anything Cloudflare is leaning towards generosity here.

Interesting fact that I wasn’t aware of. It makes sense when you think about it. The kind of people who know how to do good work at this layer level don’t have a million employers to pick from to do the most interesting work, and a few of those are now pretty gigantic (the hyperscalers) and are likely not quite what some of the alpha engineers are looking for anymore.

Basically, the intersection of small, entrepreneurial company + major worldwide internet infrastructure doesn’t have that many netizens.

Anyway, tip of the hat to Byrne for linking to this old joint in his post.

Thanks man! (and thanks for being a day-1 Extra-Deluxe supporter too!) 💚🥃

Also on the topic of Cloudflare:

Last week I installed Cloudflare for Teams (it’s free for up to 50 users, and I'm just 1...).

I can see many ways in which on-boarding could be less confusing and smoother, but generally, it's very nice.

I'm now routing traffic through WARP+ on my desktop and phone. This goes through the Argo smart-routing network — it monitors net congestion in real-time and routes to whatever is the fastest series of links between two points including Cloudflare internal network. Wheeee!

Jeff Bezos ‘it’s baked in’ x Pandemic

Here’s a crossover thought I haven’t seen yet.

Jeff Bezos quote from 2017:

“When somebody congratulates Amazon on a good quarter I say thank you. But what I’m thinking to myself is those quarterly results were actually pretty much fully baked about 3 years ago. Today I’m working on a quarter that is going to happen in 2020. Not next quarter. Next quarter for all practical purposes is done already and it has probably been done for a couple of years.”

I think something similar is going on with the pandemic, and a lot of people have trouble with both the non-linearities, delays/long-term thinking, feedback mechanisms, and reflexivity involved in the whole thing.

By that I mean that what we did (or didn’t do) early in the pandemic baked-in a lot of what we saw later. We can see the stark difference between countries that acted decisively and rapidly and those that were slow and took half measures (although there’s no perfect control group, but the logic makes sense). More pain at first saves a lot of pain later on, and is quite a good investment…

Also the difference between those with a high level of ‘everybody’s rowing in the same direction’ and those where ‘everybody is confused and half the people are going one way while the other is going in a different direction’ (ie. The U.S. and Brazil after leaders spread all kinds of BS for a long time and the epidemic became particularly politicized).

There’s also plenty of feedback loops and reflexivity; when things start going well, people relax and get sloppy, then contagion goes up, but it only shows up a numbers of weeks later… By the time it gets bad enough to scare people back into being careful, there’s many weeks of bad news baked-in, etc.

Rinse and repeat.

It’s all a big coordination problem, which is why all the forces making society less coordinated and eroding trust in those trying to coordinate the response have such a big impact. Because it’s exponential math, small differences become huge over time.

The idea isn’t that you need to follow a perfect plan or a leader who’s always right (doesn’t exist), but like wartime, you can’t have every unit go in a different direction and work at cross-purposes.

You’ll get worse results even if some of those small units do have better plans than the official one.

Crowdstrike Customer Growth

the number of customers with more than $1 million in ARR stands at 176 compared to just 10 in FY '17. This is a 105% 4-year CAGR. 

The number of customers with ARR between $1 million and $100,000 stands at 1,569 compared to just 151 in FY '17. This is an 80% 4-year CAGR.  

smaller accounts with ARR below $100,000 stand at 8,151 compared to just 286 in FY ’17. This is a 131% 4-year CAGR. 

putting this into perspective of ARR contribution, it's approximately 40% from accounts greater than $1 million in ARR, 40% from the mid-range accounts and 20% from customers less than 100,000 in ARR.

Kinda bonkers how fast these modern cloud companies can grow organically with very little capital needs and good economics (in fact, FY '21 Crowdstrike had a FCF margin of 33%, expected to come back down near term because of heavy investment, but to be 30%+ longer term; and gross margins went from the mid-30s in 2017 to almost 80% now).

Hard to imagine something similar a few decades ago. The closest was probably media and other intellectual property businesses (f.ex. Moody’s), which is why it’s an area that Buffett liked so much at the time.

‘Squarespace officially filed Friday to go public to raise some $100M’

Squarespace, Inc. today announced that it has confidentially submitted a draft registration statement on Form S-1 with the Securities and Exchange Commission.

Gotta raise some money to pay for all those podcast ads!

Seriously, a lot of good content exists thanks to them.

Half web-hosting, half media company/patron of the arts.

Video Games are Popular, March 2021 Edition

“March 2021 consumer spending across video game hardware, content, and accessories reached a March record $5.6 billion,” NPD analyst Mat Piscatella said.”[That is] 18% higher when compared to a year ago.”

“First quarter consumer spending totaled $14.9 billion, 30% higher than 2020’s first quarter,” said Piscatella. (source)

Matt Ball puts it in perspective:

In 2020, mobile gaming revenues exceeded the entire gaming market of 2013

Even though hardware is up a lot thanks to some new gen consoles, most of the money is still spent on content:

Top 5 best selling games in March:

1. Call of Duty: Black Ops Cold War

2. Monster Hunter: Rise

3. Outriders

4. Super Mario 3D World (digital sales not included, for I don’t know what reason, so it should probably be higher)

5. Marvel’s Spider-Man: Miles Morales

I hope Activision has a good R&D group looking into new variants on the word salad that is the title of Call of Duty games now, because they’ll soon run out of variants after they released the logical culmination of the series:

• Call of Duty: Heroes of Modern Ghost Black Ops Urban Warfare Strike Team World War III

Hoorah.

‘Average price of Canadian home rising at fastest annual pace ever, now up to $716,828’

The Canadian Real Estate Association said Thursday that more than 70,000 homes were sold last month, obliterating the previous record for the month by 22,000 transactions. The figure was 76 per cent higher than the same month a year ago, which saw sales slow because it was the first month of restrictions related to COVID-19.

On the price side, the average selling price for a home sold on CREA's MLS system was $716,828. That's up by 31.6 per cent in a year, and the biggest annual pace of gain on record. (Source)

Seems kinda crazy to me, but hey, I’ve thought the Canadian RE market was as high as it could get for many years, so I clearly don’t know anything about it ¯\_(ツ)_/¯

‘Britain intervenes in Nvidia’s $40 billion Arm takeover on national security concerns’

No details yet. Seems like selling ARM to a Japanese conglomerate isn’t that different from selling it to a US semiconductor company.

‘If everyone used Coinbase pro…’

Tanay Jaipuria:

Coinbase itself has different take rates for different users based on the “UI” you use (3-4% in the regular flow vs 0.5% for Coinbase pro). Users who use the basic flow pay more in fees than those who use the more sophisticated UI of Coinbase pro.

This “tax” on unsophistication is a big driver of Coinbase’s revenue. If everyone used Coinbase pro (or paid the equivalent of Coinbase Pro rate), Coinbase would make ~60%-67 less money.

---

Science & Technology

Snowflake: Data Cloud for Dummies

Not kidding with their marketing. They commissioned ebooks on the template of the “For Dummies” books to explain what they do and how customers can do various things using Snowflake tools and services.

It’s also good way to learn more about the company.

The books are on this page, including “The Data Cloud for Dummies”, “Cloud Data Engineering for Dummies” and “Cloud Data Platform for Dummies”.

2009 Talk by Nvidia's CEO Jensen Huang

Back to the future…

Great to see the vision at the time, but also to hear Jensen looking back and Nvidia’s history and strategy (the first 15 minutes go through the company’s history and main chip generations — interesting how the first two chips they designed were huge failures).

The discussion of how CUDA came to be is fascinating. The insight that to get the billions in R&D to create a whole new computing ecosystem they needed to build it on the back of a chip that had “a day job” playing games — great stuff.

It’s also cool to see Jensen talk at a more technical conference. He gets geekier than at the usual product Keynotes, and that’s great.

Another cool “vision for the future” (as of 2009) is the part around 33 minutes in when he talks about the goal of using the extra power in the new GPUs not only to make things look more beautiful through the usual things like increasing polygon count or higher-res textures or fancier shaders, but also to make environments dynamic and to all worlds to be created by simulation rather than through the “manual” work of artists (ie. artists don’t have to create every cloud, but instead you can write a cloud simulation engine that then generates them, or the behavior of liquids or smoke or fire… you create the physics and then run them).

h/t Mule

---

The Arts & History

Mural by ‘SMUG’

Human clavicles are weird, if you look at ‘em long enough. Source.

‘Amazon's Lord of the Rings to Cost $465m for Just One Season’

The Hollywood Reporter has confirmed that Amazon will spend roughly NZ$650 million – $465 million in U.S. dollars – for just the first season of the show.

That's far above previous reported estimates that pegged the fantasy drama as costing an already record-breaking $500 million for multiple seasons of the show. [...]

The figures were first released as part of as part of the New Zealand government’s Official Information Act and initially reported by the New Zealand outlet Stuff. The documents also confirmed the studio's plan is to film potentially five seasons in New Zealand – as well as possible, as-yet-unannounced spinoff series. (Source)

4 years ago:

Bezos has also demanded that Amazon make him a Game of Thrones and specifically instructed his team of top executives to begin an active search for a smash hit show to emulate Game of Thrones.

As friend-of-the-show Jerry Capital points out, looks like he may have found it (or at least, funded it).