58: S&P Global Buys IHS Markit ($44bn), Koyfin Redesign, Semiconductors Overview, Amazon Hires 2,800 Workers/Day, Grand Theft Auto V, AWS US-EAST-1 Outage, and Molecular Dynamics

"But oh well, know thyself"

Use the word “cybernetics,”' Norbert, because nobody knows what it means. This will always put you at an advantage in arguments.

—Claude Shannon (who deserves to be as widely known as Albert Einstein, for his Information Theory) in a letter to Norbert Weiner of MIT, in the 1940s.

If you want to learn more about Shannon, I recommend the book ‘A Mind at Play’ by Jimmy Soni and Rob Goodman

I wish liberal democracy had a PR department as good as monarchy when it comes to impressionable children.

My oldest kid is playing ‘Zelda: Breath of the Wild’, and he wants to please the king and save the princess so much.

Basically every old Disney movie is all about kings and princesses...

So many children's book are about the benevolent old monarch, protecting the kingdom, etc.

Hollywood needs to get on it. I get that humans have an innate tribal instinct that makes this kind of storytelling resonate, and storytelling is much easier with simple systems and clearly delineated “good guys” and “bad guys”, “us” and “them”, but sometimes it’s worth making that water flow uphill, even if it’s harder.

Investing & Business

S&P Global and IHS Markit to Merge in All-Stock Transaction ($44bn)

Huge deal this morning.

all-stock transaction which values IHS Markit at an enterprise value of $44 billion, including $4.8 billion of net debt [...]

Upon completion of the transaction, current S&P Global shareholders will own approximately 67.75% of the combined company on a fully diluted basis, while IHS Markit shareholders will own approximately 32.25%.

Some highlights of what is expected to come from this:

The pro forma company will have 76% recurring revenue and expects to realize 6.5-8.0% annual organic revenue growth in 2022 and 2023 [...]

target 200 basis points of annual EBITA margin expansion. [...]

expects to deliver annual run-rate cost synergies of approximately $480 million, with approximately $390 million of those expected by the end of the second year post-closing, and $350 million in run-rate revenue synergies for an expected total run-rate EBITA impact of approximately $680 million by the end of the fifth full year after closing. [...]

expects to generate annual free cash flow exceeding $5 billion by 2023 [...]

Expecting adjusted DEPS in “mid-teens” by 2023

There’s worry that they’re diluting the best parts of the business (CRA, indexes), but management is quite savvy and I would tend to give them the benefit of the doubt. They could over-deliver… It does seem to me like they could’ve used more debt, being at 0.4x net-debt-to-EBITDA. Release. Slides.

Koyfin Major Redesign (Lot of New Features + Refreshed UI)

This is big enough that I don’t have time to write about it all today, but definitely check it out, if you don’t already constantly have the site open in a tab like I do.

But a few things that will make your life easier:

  • You now need to hit “/’ before keyboard commands

  • Try out the sidebar that can pop out from the right (3 options, near the top right). I particularly like how you can see your watchlists there while hanging out in the “HOME” view. Now that’s a true command center!

  • If you click on your account icon on the top right, you can switch the color theme. I’m undecided between ‘Monochrome” and “Pitch black” on this version… I’ll give each a few days and see where my brain settles. But I think I prefer “Monochrome” so far.

Check out the release notes here.

Semiconductor Industry Overview with Jon Bathgate and Brinton Johns

Excellent conversation between Shane Parrish (hey Shane!) and Jon and Brinton from NZS Capital on the semiconductor industry, with a quick overview of the major players (Intel, AMD, Nvidia, TSMC, Samsung, ASML, Cadence, etc), some of the geopolitics involved (Taiwan is the linchpin of most of it), and a little bit on analog semis at the end (Texas Instruments’ long-duration, wide-catalogue approach).

I love how Shane seems to casually knowns the wavelength of various segments of the light spectrum at one point (he may have had that chart in front of him, but it still sounds badass).

This is good stuff, especially if you’re not super familiar with semiconductors and want a high-level primer of the major moving pieces and what to keep an eye on.

If you’re a Farnam Street Member, you can access the podcast version and the transcript and show notes here:

And there’s also the Youtube version that is publicly available to all (see above).

Also, I discuss both Nvidia and Texas Instruments a bit in this interview I did with Rob of Koyfin (some context about it in edition #46).

Are You Working for Amazon Yet? (2,800 hires a day)

Seems like if we extrapolate the current trend a bit, we’ll soon all be working for them:

Amazon has embarked on an extraordinary hiring binge this year, vacuuming up an average of 1,400 new workers a day [...]

Amazon added 427,300 employees between January and October, pushing its work force to more than 1.2 million people globally, up more than 50 percent from a year ago. [...]

Starting in July, the company brought on about 350,000 employees, or 2,800 a day. [...]

The scale of hiring is even larger than it may seem because the numbers do not account for employee churn, nor do they include the 100,000 temporary workers who have been recruited for the holiday shopping season. They also do not include what internal documents show as roughly 500,000 delivery drivers, who are contractors and not direct Amazon employees.

What’s the historical comparison? Shipbuilding in WWII:

Such rapid growth is unrivaled in the history of corporate America. It far outstrips the 230,000 employees that Walmart, the largest private employer with more than 2.2 million workers, added in a single year two decades ago. The closest comparisons are the hiring that entire industries carried out in wartime, such as shipbuilding during the early years of World War II


Interview: Dan McMurtrie a.k.a. @SuperMugatu (Top 5 of 2020)

Ok, I’m calling it now. This podcast interview of Dan McMurtrie by Bill Brewster is a Top 5 financial podcast for 2020. Possibly Top 3, but that would require longer thought…

Frankly, the main downside of this is that it has made me a little jealous. I wish I was that smart about investing and came at things from 7 different angles using 11 different frameworks and hundreds of interviews and conversations that recursively iterated on themselves.

But oh well, know thyself ¯\_(ツ)_/¯

‘7 Days Out: Eleven Madison Park’…

There’s a Netflix documentary series called ‘7 Days Out’ that looks at the last-minute behind-the-scenes preparations required for various events (a fashion show, a League of Legends tournament, etc). Most episodes aren’t that great, but there’s one that really stands out, the one about the re-opening of the world-renowned restaurant in NYC ‘Eleven Madison Park’:

…and More Nick Kokonas

Speaking of great restaurants, if you enjoyed the ‘Chef’s Table’ (Netflix) episode about the Alinea and Patrick’s interview with Nick Kokonas about his restaurant experience, I think you may also enjoy this one. It’s also about a chef-businessman partnership (though with a different flavor), and both the culinary and business aspects of it are interesting.

What stands out from both stories, though, is the relentless pursuit of perfection and the attention to the most minute details. If you want to be at the top, nothing less will do.

And here’s a bonus: I found out that Nick Kokonas did an interview with Tim Ferriss a few months ago. He has a great business mind and it’s great to see him describe how he thought about the pandemic before it hit the US, and the wartime mobilization of resources and ingenuity that he marshalled after it hit.

He did another one back in 2018 that is three hours long and goes into his career as an options trader in Chicago, and a great deconstruction of the publishing industry, which may interest you. You can follow Nick on Twitter here.

Science & Technology

‘Pushing the Limit of Molecular Dynamics with Ab
Accuracy to 100 Million Atoms with
Machine Learning’

Ok, this is cool. I wish I understood more of it, but I’ve been interested by computational biology for a while, following projects like Rosetta@home from the Barker Lab and Folding@home at Stanford, and the CASP competitions every couple years.

DeepMind has recently showed some really interesting work with AlphaFold (came out on the scene in 2018) and there’s always a bunch of cool stuff happening.

This paper seems to be the latest step forward:

For 35 years, ab initio molecular dynamics (AIMD) has been the method of choice for modeling complex atomistic phenomena from first principles. However, most AIMD applications are limited by computational cost to systems with thousands of atoms at most. We report that a machine learning based simulation protocol (Deep Potential Molecular Dynamics), while retaining ab initio accuracy, can simulate more than 1 nanosecond-long trajectory of over 100 million atoms per day, using a highly optimized code (GPU DeePMD-kit) on the Summit supercomputer. Our code can efficiently scale up to the entire Summit supercomputer, attaining 91 PFLOPS in double precision (45.5% of the peak) and 162/275 PFLOPS in mixed-single/half precision. The great accomplishment of this work is that it opens the door to simulating unprecedented size and time scales with ab initio accuracy. It also poses new challenges to the next-generation supercomputer for a better integration of machine learning and physical modeling.

This is such a hard problem because as you increase the complexity of the molecule(s) you’re trying to simulate and the timescale, computational needs explode non-linearly:

The computational cost of AIMD generally scales cubically with respect to the number of electronic degrees of freedom. On a desktop workstation, the typical spatial and temporal scales achievable by AIMD are ∼100 atoms and ∼10 picoseconds. From 2006 to 2019, the peak performance of the world’s fastest supercomputer has increased about 550-folds, (from 360 TFLOPS of BlueGene/L to 200 PFLOPS of Summit), but the accessible system size has only increased 8 times

This last sentence is why we need to find ways to accelerate and/or cut down on the branches of the possibility tree that are being looked at during these simulations.

Why do we need to be able to simulate these bigger and longer reactions?

For problems in complex chemical reactions electrochemical cells, nanocrystalline materials, radiation damage, dynamic fracture, and crack propagation, etc., the required system size typically ranges from thousands to hundreds of millions of atoms. Some of these problems demand time scales extending up to the microsecond and beyond, which is far out of the scope of AIMD

Hundreds of millions of atoms over microseconds! If that level of complexity doesn’t both totally blank out your mind and make you go 🤯, you need to think about it for a little longer.

Anyway, this is as far as I’ll go about this here, not sure how many computational biology fans there are in the audience, but you can read more here.

AWS Post-Mortem on US-EAST-1 Outage

Interesting if you enjoy looking behind the curtain at how the various pieces of internet infrastructure work together:

At 9:39 AM PST, we were able to confirm a root cause, and it turned out this wasn’t driven by memory pressure. Rather, the new capacity had caused all of the servers in the fleet to exceed the maximum number of threads allowed by an operating system configuration. As this limit was being exceeded, cache construction was failing to complete and front-end servers were ending up with useless shard-maps that left them unable to route requests to back-end clusters. [...]

The front-end fleet is composed of many thousands of servers, and for the reasons described earlier, we could only add servers at the rate of a few hundred per hour. We continued to slowly add traffic to the front-end fleet with the Kinesis error rate steadily dropping from noon onward. Kinesis fully returned to normal at 10:23 PM PST.

Reading this reminds me a bit of the last episode of Chernobyl when they explain moment by moment what happened (with much lower stakes, of course)…

It’s interesting to see how many things are interdependent and how even a system designed for reliability can always find a way to prove Murphy’s Law right.

‘The story of mRNA: How a once-dismissed idea became a leading technology in the Covid vaccine race’

We’re going to hear a lot about mRNA vaccines in coming months, but how did we get here?

Sometimes science looks like this big amorphous force that just makes things happen automatically, but there are real people grinding it out and battling demons for every millimeter of forward movement:

“Every night I was working: grant, grant, grant,” Karikó remembered, referring to her efforts to obtain funding. “And it came back always no, no, no.”

By 1995, after six years on the faculty at the University of Pennsylvania, Karikó got demoted. She had been on the path to full professorship, but with no money coming in to support her work on mRNA, her bosses saw no point in pressing on.
She was back to the lower rungs of the scientific academy.

“Usually, at that point, people just say goodbye and leave because it’s so horrible,” Karikó said.

There’s no opportune time for demotion, but 1995 had already been uncommonly difficult. Karikó had recently endured a cancer scare, and her husband was stuck in Hungary sorting out a visa issue. Now the work to which she’d devoted countless hours was slipping through her fingers.

“I thought of going somewhere else, or doing something else,” Karikó said. “I also thought maybe I’m not good enough, not smart enough. I tried to imagine: Everything is here, and I just have to do better experiments.

Never give up, never surrender.

Source. h/t Patrick Collison

‘Starting the computer like it was one of those old lawnmowers’

Good old-school computing anecdote from a comment on Hacker News:

When I was a CS major in the 90's, one of my professors told me a story of his own college days, with punch-card computers.

His university bought a tape reader (like, punched paper tape, not magnetic tape) to do the boot code of the computer, on the theory that tape was a little easier to manage than punch-cards for the boot (you can't lose one of the cards, or get them out of order, etc with tape). So my prof and some of his friends start playing with the tape reader, and they realize that what controls the IO speed of the tape is actually the tensile strength of the tape -- if the feeder tries to put too much force on it, it will tear the paper tape. The actual computer can read the instructions much faster than the tape can physically handle.

So they got some plastic tape instead, and punched the boot code in the (much stronger) plastic tape. Then, to boot the computer, they'd feed the plastic tape through the part of the reader that actually read, bypassing the mechanical part that pulled and wound the tape, and then manually grab the other end and yank on it as hard as they could, basically starting the computer like it was one of those old lawnmowers that you pulled the cord to turn over the engine.

Does this count as over-clocking?

‘Grand Theft Auto V Has Outlasted An Entire Console Generation’

Matt Ball shared this piece from Kotaku about GTA V with this commentary:

• Built for 7th Gen (where it was the best-selling title)
• Launched on 8th (also best-selling title)
• Kicked of Sony's 9th gen presentation
• Single highest rev media title ever
GTA:V Has Outlasted An Entire Console Generation

[Cumulative revenue has been:] $7-8B


Grand Theft Auto V, in case you forget, was first released on the Xbox 360 and PS3 in September 2013. It didn’t arrive on the PS4 and Xbox One until 2014. [...]

Grand Theft Auto V’s slow transformation from a single-player experience to a microtransaction-riddled online playground is one of the starkest examples of the way the video game industry has changed how it treats its product, and customers, in the PS4/Xbox One years.

In Theory, Theory and Practice are the Same. In Practice…

The Arts & History

Computer, define 'dancing.'

On a movie night with the kids recently, we watched 'Wall-E' (2008, Pixar). Classic stuff.

Eve is so badass.

Trivia: Jony Ive of Apple helped design her. She does look like an Apple product.

Wall-E director Andrew Stanton said he wanted the design to be high-end, but also "seamless and for the technology to be sort of hidden and subcutaneous." Stanton called this philosophy straight out of the Apple playbook and called up El Jobso [Steve Jobs] in 2005. Steve sent over Johnny Ive for the day. (Source)

Not to nitpick a great film, but isn't it kind of a plot-hole that the plant was growing inside a fridge (or at least, behind a door)? Seems like it wouldn't get enough light...

It doesn't matter much. I just wonder why they made that choice. They probably wanted to set up the laser cutting tool for later, when Wall-E needs it to escape the broken-robot prison, but they could’ve done that some other way...

‘Chess sets become hot holiday gift’

Nice to see a high-quality show have a second order positive effect on culture this way:

Chess-related items are on the holiday wish list this year, thanks to the popularity of Netflix’s hit new show “The Queen’s Gambit.” Sales of chess items were up by 300% on Thanksgiving compared with the previous month, according to data from Adobe Analytics. The company tracks transactions of 80 of the 100 largest retailers in the U.S. (Source)

Here I gotta do the annoying name-dropping thing, but I was excited to see that my tweet about this got retweeted by Garry-freaking-Kasparov!

I’ve long admired him for his chess mind, but now I equally admire him for his defense of liberty (I didn’t pick this name for nothing) and his anti-authoritarian advocacy.

I finished watching the last episode of ‘Queen’s Gambit’ last night. I’ll be writing more thoughts about it in a future edition. But short version: I recommend it.