144: Jim Killer Epic Interview, Defense Mergers 1980-2001, MSCI 42% CAGR, TSMC, Cloudflare's Clever Black Hat Strategy, (De/In)flation, Russia & Cyberattacks, and Art Talent

"maybe a Dyson sphere"

It is impossible for a person to begin to learn what he thinks he already knows.

—Epictetus, Discourses, 2.17.1

💵 ⏳ As an investor, you have to enjoy the process for it to make sense. If you don’t, go do something else.

Why?

Because the process takes so long to work (unless you basically win the investing lottery, and you shouldn’t count on that) that you spend decades of your life in it before the exponential curve really blows your mind.

You don’t want to be the guy in his early 30s who can’t wait to be 60 just so compounding can have had a chance to really work.

It’s like that ridiculous film with Adam Sandler where he gets a magical remote that allows him to fast-forward the boring parts of his life, and he ends up forwarding through what turned out to have been all the good parts because he’s always living for something in the future.

Don’t live for tomorrow.

Plan for tomorrow. Set things up so that tomorrow’s good. Present-me is always thinking of how he can help future-me…

But you gotta enjoy the now, because it’s the only thing that is real. 🗿

🐦 Twitter has to have more volunteer product designers than any other company out there… So what’s one more, right?

But I’ll share my ideas anyway:

It should be possible when you click on someone’s profile and then click on who they follow to see a timeline view based on these people. In other words, you should be able to see the the timeline feed that someone else sees.

Everybody’s Twitter is different, so it would be really powerful to be able to dip in each other’s reality and see what it’s like.

And if you like what you see, you should be able to save it as a list on your own profile for convenient access. It would help tremendously with onboarding, some people would become known as great curators of follows.

💪 I have a pull-up bar screwed to a doorframe about 10 feet from my desk (this exact model — but use the screws, don’t do the pressure fit, you’ll end up on the floor eventually).

It’s all about reducing friction, increasing convenience. I’m not one of those people who love to exercise and are constantly at the gym, so it has to be in my face and easy.

A few years ago it worked pretty well. I had given myself the challenge of being able to do 10 full pull-ups before the end of that year, and ended being able to do a little above 10 (I don’t remember exactly.. somewhere in the 11-13 range).

I was pretty happy with that and saw how good an exercise it was at strengthening my back and arms (I read the book ‘Convict Conditioning’, a great book about bodyweight exercises written by someone who spent a looong time in prison without much equipment… the ways to progressively increase resistance just by adjusting technique/angles/limb isolation).

Anyway, the problem is that after reaching that goal, we had a second kid, then the pandemic hit, it all messed up with my routine/discipline/willpower, and I lost most of my capacity to do pull-ups. I can do about 4 now, maybe 5 if I give it my all.

I’ve set myself the goal of getting back to 10 (and then I won’t rest of my laurels, I’ll do another goal).

If you want to join me, it’s easy, you just need a bar that you can grab and lift yourself from, and you need to figure out what is a goal that is doable, but hard, and will required a few months of consistent work to reach. For you it could be 1 full pull-up or it could be 100, it’s pointless to compare between people — just compare with your own potential.

Let me know in the comments if you join, the public commitment will help you do it (psych 101).

📚 I’ve never owned a Kindle. But I think that era is over.

On Prime Day, I’m buying a Kindle Paperwhite.

I want to reduce friction as much as possible between me and books, and having one more lightweight and practical option to read can’t hurt.

👾 This is kind of a follow-up to my time-travel musings at the beginning of edition #140.

It’s an idea for a science fiction short-story that I came up with on a walk (talked it out a bit with my friend JPV too).

So the characters are some sorts of special ops team, delta force or Navy SEALs types.

There’s some huge conflict brewing in what looks to be the near future. It’s the edge of nuclear war.

You learn that some kind of secret Manhattan project has created a teleportation device, but it has been kept secret so far so that it can be used to win this exact kind of high-stakes war through the element of surprise.

The plan is to teleport the strike team deep into the command bunker of the adversary country and do a decapitation move, to kill off the leaders of the regime, who have the launch codes and are unstable and suicidal for some reason (f.ex. they know they’re going to lose the war, and rather than be captured or killed, they’d rather take everyone with them via global thermonuclear war). But if they can be removed, chances of nuclear war go down by a lot…

They use the machine to teleport, and when they re-materialize there, things don’t look at all like they expect. Their electronic guidance can’t get a signal and their satellite comms don’t work, so they’re on their own.

As things progress, you realize they’re in an alien world where a much more advanced civilization lives. I haven’t really thought about the exact tech tree and how I’d describe the life-forms, but I’m mostly thinking of something with AGI, nanotech, geoengineering, maybe planetary-size projects like large space habitats, maybe even a Dyson sphere, etc.

So the team starts to speculate… Did the device malfunction, did we get beamed to another planet? What are the odds? Space is mostly empty, how did we end up on a planet where we can breathe the air and where there’s advanced life?

Did we go to another dimension? Is that even possible?

So my big twist near the end is… [This will be revealed only to paid supporters] 💚🥃

I’m kidding, kidding.

My big twist is that through some weird quirk of space and time being related, the teleportation machine actually sent them through both time and space, so they’re still on Earth, but hundreds of years in the future in a very different world (things change faster than we expect if we think linearly; but if progress is building on past progress and speeding up, a few hundred years will be plenty for really big changes).

💚 🥃 I figure that the price of a couple coffees or one alcoholic drink isn't a bad trade for 12 emails per month full of eclectic highlights and investing/tech analysis.

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).

As Bezos would say of Prime, you’d be downright irresponsible not to be a member, it takes 19 seconds (3 on mobile with Apple or Google Pay):

---

Investing & Business

Share of Retail that is Online

This is a slide form Stripe’s keynote, via Mario Gabriele.

Clever Way Cloudflare Learns from Bad Guys…

I’ve written multiple times about how one of Cloudflare’s main assets is the network itself, and how large it is because of the tens of millions of free users.

The company talks about how this acts both as a very good way to rapidly release and test new products, with basically infinite beta-testers and lots of feedback to iterate products and uncover bugs. They also describe it as an incredibly unique sensory organ that englobes a large part of the internet and can discover patterns that are then used to build products (for security, performance, reliability, etc).

But there’s an aspect that I haven’t seen the company mention, probably because it doesn’t make for a great PR moment, but it’s still very cool.

If you have a free product, anyone can sign up. Among the free users, there will be black hats attempting hacks and DDoS (distributed denial-of-service attacks, basically trying to send so much traffic to a site from a variety of sources that it overwhelms the target and it becomes inaccessible to real users).

Cloudflare can watch these assholes on its own network and learn from how they conduct their business how to better counter their actions.

In other words, the Cloudflare network acts as a sensor on both the receiving and the attacking side for many types of attacks, and that gives it better data to create effective antidotes.

The beauty is, I don’t think anyone else is really doing this, because others don’t have Cloudflare’s very large free userbase. Some other types of security-related companies, like large VPN providers, no doubt have plenty of black hats on their networks, but they don’t build security products, so what they learn is of limited use (they can observe patterns and try to kick malicious users out).

h/t to reader C.B.

Live chart on Koyfin

MSCI 42% CAGR since 2012

If you cherry-pick the starting point a bit, MSCI have a stretch since November 2012 where they did over 42% CAGR over 8.58 years.

And they did it looking ‘expensive’ all the way up.

I wonder what performance S&P Global’s equivalent segment would’ve had over the same period if it had been a public company. Pat Srinivas dug up the index revenue over time vs MSCI and they look pretty similar.

h/t Friend-of-the-show Jerry Capital (💎 🐶)

Defense Mergers, 1980-2001

Source (p.134)

h/t Fat Tail Capital, Emma Salisbury

Inflation? Deflation?

I liked this Tweet by Extra-Deluxe supporter (💚💚💚💚💚 🥃) John Mihaljevic:

The real answer might be that we have inflation *and* deflation. Deflation in all things bits, inflation in all things atoms. Bottom line: Unless you live in the world of bits, you better figure out how you'll pay for the things you desire in the physical world. Or, desire less.

---

Science & Technology

Epic Interview: Jim Keller (on his career at AMD, Intel, Tesla, Apple, Tenstorrent, AI chips vs GPUs, etc)

Or should I say ‘epyc’ interview? 😬

The silicon OG is always interesting :

• AnandTech Interview with Jim Keller: 'The Laziest Person at Tesla'

I'm not the guy to tweak things in production – it’s either a clean piece of paper or a complete disaster. That seems to be the things I do best at. It's good to know yourself - I'm not an operational manager. So Tenstorrent is more the clean piece of paper. The AI space is exploding. 

Know thyself.

Sure, mitigate your flaws, but mostly, try to accentuate and build on your strengths.

people aren't getting any smarter. So people can't continue to work across more and more things - that's just dumb. But you do have to build tools and organizations that support people's ability to do complicated things. The VAX 8800 [DEC computer from 1984] team was 150 people. But the team that built the first or second processor at Apple [the A-series chips, around 2010-2011], the first big custom core, was 150 people. Now, the CAD tools are unbelievably better, and we use 1000s of computers to do simulations, plus we have tools that could place and route 2 million gates versus 200. So something has changed radically, but the number of people an engineer might talk to in a given day didn't change at all. If you have an engineer talk to more than five people a day, they'll lose their mind. So, some things are really constant.

I love this. I often think about how people 10,000 years ago weren’t dumber than we are, at least not in the raw abilities of their brains. It’s all the mind-software that we have today that makes us more effective, so it’s at this level that we have to innovate.

Keller seems to think that there’s a better way to do AI than with GPUs (even if those GPUs get modified over time to be more specialized for AI and not just for graphics), and that’s what he’s working on at Tenstorrent:

There are genuinely three different kinds of computers: CPUs, GPUs, and AI. NVIDIA is kind of doing the ‘inbetweener’ thing where they're using a GPU to run AI, and they're trying to enhance it. Some of that is obviously working pretty well, and some of it is obviously fairly complicated. [...]

GPUs were built to run shader programs on pixels, so if you're given 8 million pixels, and the big GPUs now have 6000 threads, you can cover all the pixels with each one of them running 1000 programs per frame. But it's sort of like an army of ants carrying around grains of sand, whereas big AI computers, they have really big matrix multipliers. They like a much smaller number of threads that do a lot more math because the problem is inherently big. Whereas the shader problem was that the problems were inherently small because there are so many pixels.

The sculptors’s approach to team-building, very interesting stuff:

The first thing you have to realize is if you are building the team, or finding one. So there's a great museum in Venice, the David Museum, and the front of the museum, there's these huge blocks of marble. 20 by 20 by 20. How they move them, I don't know. The block of marble sitting there, and Michelangelo could see this beautiful sculpture in it. It was already there, right? The problem was removing the excess marble.

So if you go into companies with 1000 employees, I guarantee you, there's a good team there. You don't have to hire anybody. When I was at AMD, I hardly hired anybody. We moved people around, we re-deployed people [elsewhere], but there were plenty of great people there. When I went to Tesla, we had to build the team from scratch, because there was nobody at Tesla that was building chips

There’s plenty more interesting stuff that I don’t have the space to include here. It’s a long and geeky interview, but I love that stuff, and maybe you do too.

h/t Friend-of-the-show and supporter (💚 🥃) Mostly Borrowed Ideas

TSMC: Relentless

TSMC is fast advancing its manufacturing processes, and it is expected to move its N4, namely 4nm, node to risk production in the third quarter of 2021, with N3 - 3nm node - to start volume production at the world's number-one pure-play foundry house in second-half 2022. [...]

The advanced N3 chip tech uses the FinFET transistor architecture. TSMC says N3 will offer up to 15% faster speeds or consume 30% less power than N5 with an up to 70% logic density gain. (Source)

Biden gives Putin list of 16 “off-limit” areas for cyberattacks

U.S. President Joe Biden told Russian President Vladimir Putin on Wednesday that certain critical infrastructure should be "off-limits" to cyberattacks [...]

Biden wasn't explicit about which areas he wanted out of bounds, but spoke of 16 kinds of infrastructure - an apparent reference to the 16 sectors designated as critical by the U.S. Homeland Security Department, including telecommunications, healthcare, food and energy.

"We agreed to task experts in both our countries to work on specific understandings about what is off-limits," Biden said following a lakeside summit with Putin in Geneva. "We'll find out whether we have a cybersecurity arrangement that begins to bring some order." (Source)

This is really easy to make fun of. The jokes write themselves.

“Oh, so everything else is acceptable? Did Biden just endorse Russian cyber-attacks on large swaths of the country? Who do lobbyists have to convince to get their industry on the list? etc”

But in realpolitiks, you can't just say "everything is off-limits, red line everywhere, anything you do will be responded to with equal force", because you know you won’t go to war over a 7-11 being hacked, and the more you don’t follow through on threats, the less credible they because… so I guess it's a pragmatic step in the right direction, and better than what the U.S. had before.

It remains to be seen how well both the defensive and deterrence moves that the US is making on cybersecurity will work, and how long it’ll take for measures to be implemented.

In theory, the zero trust world that we’re moving towards should be better than the moat & castle world of yesterday, but as always, the devil is in the details, and in theory, theory and practice are the same, but in practice, they aren’t.

---

The Arts & History

‘20-minute sketch’

Here it is:

• a quick 20min lunchtime sketch ☺️

The artist is Julie Hilboldt, you can see more of her work here..

I don’t know what that style is called. I don’t know what this teardrop-brush is called.

But I really like it.

I showed it to my wife, and we were talking about how impressed we are that some people can become so good at art, and how it seems almost impossible to be that good, at least to us who don’t have these types of skills.

But it’s also a good reminder of the power of compounding (“time and pressure”, as Andy Dufresne would say).

We have to remember how long they’ve been training, and how all these little improvements month-by-month, year-by-year compound into this level of skill.

They probably started drawing stick figures just like I did, but I did them once in a while and never really did much deliberate practice.

Artists like my cousin, they were drawing and painting tens of hours a week from a young age, and never stopped. Took classes, read art books, and compared notes with other artists in ways specifically designed to improve and explore new techniques.

It doesn’t just happen.

It’s true that we don’t all start from the same place — some people are more naturally talented — but it doesn’t mean what most people think.

Some of the best artists we see may not be some of the most naturally talented, and some failures are no doubt terribly gifted. It takes a lot more than that, we shouldn’t underestimate deliberate practice and number of reps.

👋