348: Tacit Knowledge, Charter, Stripe Layoffs, Hollywood, Meta's Computational Biology AI, Floppy Disks, and Age Segregation
"It’s a blueprint."
I see my age in my children. —Dave Chapelle
🚶🏻♂️🚶 Yesterday, I spent a couple of hours walking along the river and downtown Ottawa with friend-of-the-show Trevor Scott. It was fun! I so rarely see in-person my online-first friends, it was a good reminder that there are real people behind the avatars and usernames. 👨💻
📝🎙🎯 Follow-up on Ben Thompson’s slow pivot from edition #347:
A few smart people (thanks Trung, Evie, Ant) pointed out that Bill Simmons’ evolution is a good analog for what Ben has been doing.
I think it’s a bullseye. I didn’t see it because I’m not a sports person, so Simmons only really got on my radar after his transformation (with non-sports interviews, podcasts like the Rewatchables, etc).
But knowing how Ben has been following Simmons’ career, how they’re friends (I think — they sure seem friendly on the interviews they’ve done together), it’s not surprising at all that there may have been an influence there, conscious or not.
The other important aspect that I didn’t cover, and that the man himself pointed out to me, is the custom-built software layer that makes it all possible. It allows the bundling of multiple shows, subs can select exactly what they want, and it provides authentication/security for paid accounts (and who knows what features are coming next).
It may not seem like much from the outside, but trust me, it’s a lot of work to build *and* immediately scale to tens of thousands of users (I’m guessing).
🧓🏻👴🏻 ⚖️ For a long time, I’ve been meaning to write about how segregated our society is by age, at least in North America, where I live.
Once in a while, I try to think about how often I have a real conversation longer than 5 minutes with someone who’s at least 15 years older or younger than I am who’s not extended family or that I’m not doing some transactional commercial/job-related interaction with.
I don’t know about you, but for me it’s surprisingly rare, and yet I suspect I have more of those than most people because I spend my days on the internet where age matters less, and I’ve made friends who are much younger and older than I am.
I’m not sure I understand all the factors that created this dynamic, but the end result is undeniable.
I also find it sad that ageism is one of the last socially acceptable (or at least mostly invisible) ways to discriminate against whole groups of very different individuals based on some visible part of who they are that they didn’t choose.
Many people may look at this chart and assume that a lot of the problems in US politics are due to the age of representatives, because being old is so widely seen as a negative by so much of the population that it’s the immediate explanation that jumps to mind.
Maybe instead we could see the rising age is a *symptom* of dysfunctions in the system that gives such a big advantage to incumbents (name recognition, gerrymandering, fundraising advantages, accumulated political favors, etc) that they tend to stick around forever.
Or that politics has become so unattractive to most high-quality people that those who end up interested in it are exactly the people who you wouldn’t want to see run things and have power?
Among every age cohort, young and old, there are a bunch of idiots and geniuses, honest and dishonest people, learning machines and calcified fossils who haven’t read anything longer than the back of a box of cereals in years, etc.
I’d rather have a bunch of old, hard-working, smart, honest engineers, entrepreneurs, artists, scientists, doctors, historians, well-respected military leaders, etc, than a bunch of young power-hungry ideological narcissistic lawyers and 3rd-rate partisan think-tank consultants.
Ideally, you’d have a mix of quality people from a diverse range of backgrounds, skills, and age cohorts (because that shapes their life experience). To get that, you need to make politics attractive to these people — but how does politics sound as a career choice to you (because I know you’re a quality person that I’d be happy to see run stuff)?
Yikes, uh? 😬
Well, there it is.
💚 🥃 🧠 Are you reading this? I guess you are. You have to be, or these words wouldn’t be in your brain right now. Thank you for trusting me with your brain!
Liberty’s Highlights is reader-supported. To support my work, consider becoming a paid supporter. 🧠
🧩 A Word From Our Sponsor: Heyday 🧠
Do you have 100+ browser tabs open right now? 😬
Give your memory a boost with Heyday, the research tool that automatically saves the content you view, and resurfaces it within your existing workflows. 👩💻
It’s like cheat codes for your memory. 😲💡
Business & Investing
🌐 Large scale global Tacit knowledge transfer via.. YouTube 🏀🤺🏌️♂️🎤🎸🍳🧑🏻🍳⏲🎥🎬🎨👩🎨
Tacit knowledge is extremely important, yet I don’t think we’re very good at managing it, transferring it, or even sometimes remembering how crucial it is (from the basketball court to TSMC fabs).
Tacit knowledge or implicit knowledge—as opposed to formal, codified or explicit knowledge—is knowledge that is difficult to express or extract, and thus more difficult to transfer to others by means of writing it down or verbalizing it. This can include personal wisdom, experience, insight, and intuition.
[...] the ability to speak a language, ride a bicycle, knead dough, play a musical instrument, or design and use complex equipment requires all sorts of knowledge which is not always known explicitly, even by expert practitioners, and which is difficult or impossible to explicitly transfer to other people.
Friend-of-the-show and supporter (💚 🥃) Trung Phan has a great post about the power of YouTube to help record, preserve, and spread a lot of tacit knowledge. Worth reading in full.
It reminded me of a recent podcast by friend-of-the-show David Senra (🎙📚) about how Kobe Bryant would spend all his free time (when he wasn’t training) watching VHS tapes of Michael Jordan and other greats, sometimes from decades before his time, imprinting on their technique.
This pattern repeats because it’s not just a “hmm, interesting” bit of trivia.
It’s a blueprint.
If you want to get good at something — really good — you need to study those that came before you, learn what took them decades to learn in a compressed period, learn from their mistakes, and try to build on top of it using your own personal strengths and weaknesses and unique attributes.
It doesn’t mean you’ll be great — but you’ll certainly be greater than if you don’t do a deep study of the past and try to reinvent every wheel yourself. GOOG 0.00%↑
🚠 Tom Rutledge’s Exit Interview
When you think about what the average broadband bill is, in our company with promotions and everything else, our average revenue per customer is about $64.
The average mobile customer inside our footprint is spending about $135 a month on mobile service — multiple lines through for all members of the household.
So there’s a lot more money being spent on mobile than there is on broadband. And yet broadband is a significantly richer product from a data throughput perspective. And we can actually make the mobile product, which is used 85% of the time in the home or in the office and on the Wi-Fi system, we can make that an even faster service in the home and in the office, and we can make it a less expensive service.
I remember when we launched the triple play for wireline, data and video, the average phone bill in the New York metropolitan area was about $78 [per month]. We brought that down to $30 and ended up having the majority of the customers. I think we have the same opportunity in mobile.
Mobile, yes, is a fully penetrated business in the country, not growing that fast, but if you look at where we are in mobile, we’re not well penetrated. And so we’ve got tremendous upside for years to come.
On broadband growth:
Q: do you expect broadband growth to look anything like what it’s looked to the past, say, five, seven or nine years?
I think when you aggregate it all up, it’s got the potential to be like that. Yes. That’s still reasonable.
Q: In other words, what we’ve seen this past year is a blip between pandemic pull-through effects and macroeconomic difficulty?
That’s my view. I mean, obviously, as you reach full penetration, you’re going to have some slowing down in growth. At some point, it gets to the household growth rate. But I don’t see that for five years or more. I think there’s continuous opportunity. I do think if you look at the trend lines, 2020 was a massive blip in terms of growth and even 2021 had growth associated with the pandemic that pulled forward a lot of growth.
On fiber overbuilds:
yes, there’s been fiber expansion, although it hasn’t really changed much over the last 10 years. The pace of that hasn’t changed much over the last 10 years, even notwithstanding all the announcements that have been made recently.
It takes time to build out infrastructure. It’s very expensive. All of those who’ve done it in the past have failed. You know, if you look at Verizon’s FiOS, they ended up selling most of it. Almost all overbuilders of physical infrastructure don’t do well in the long term. So I think the macroeconomic forces that have always affected overbuilders will continue to affect them and affect the pace of construction.
On mobile & cable merging:
Q: At some point in the next five, 10 years, will we have merged wireless cable companies in this country?
Uh, you know, yes, I do think that. Some of the assets that are in each of those defined companies now will be in other companies.
There’s more, but I’ll leave it here.
14% cut. Some highlights from Patrick Collison’s letter to employees:
Today we’re announcing the hardest change we have had to make at Stripe to date. We’re reducing the size of our team by around 14% and saying goodbye to many talented Stripes in the process. If you are among those impacted, you will receive a notification email within the next 15 minutes.
What they’re seeing macro-wise:
At the outset of the pandemic in 2020, the world rotated overnight towards e-commerce. We witnessed significantly higher growth rates over the course of 2020 and 2021 compared to what we had seen previously.
The world is now shifting again. We are facing stubborn inflation, energy shocks, higher interest rates, reduced investment budgets, and sparser startup funding. [...]
We think that 2022 represents the beginning of a different economic climate.
Our business is fundamentally well-positioned to weather harsh circumstances. We provide an important foundation to our customers and Stripe is not a discretionary service that customers turn off if budget is squeezed. However, we do need to match the pace of our investments with the realities around us.
Mea culpa and other cost-cutting:
you might reasonably wonder whether Stripe’s leadership made some errors of judgment. We’d go further than that. In our view, we made two very consequential mistakes, and we want to highlight them here since they’re important:
We were much too optimistic about the internet economy’s near-term growth in 2022 and 2023 and underestimated both the likelihood and impact of a broader slowdown.
We grew operating costs too quickly. Buoyed by the success we’re seeing in some of our new product areas, we allowed coordination costs to grow and operational inefficiencies to seep in.
We are going to correct these mistakes. So, in addition to the headcount changes described above, we are firmly reining in all other sources of cost.
I feel much empathy for those affected (not just at Stripe), losing a job — especially if you really liked it — can be one of life’s great shocks. Gotta keep moving forward, though. Never give up, never surrender.
🎞📽🍿Going to the movies
Matt Ball (👨🚒):
And if you showed this chart to any film exec from 1920-1950, they’d doubtlessly assume it reflected the cataclysmic end of the medium altogether, when in fact outputs are up over 100x and revenues even more
[NB: the figures have fallen another 60% since!]
Note the last line, and how the charts stop in 1990. The lines go down another 60% since! 😯
Science & Technology
🏎 🧬 ‘Meta AI predicts shape of 600 million proteins’ 🧬
Computational protein prediction and design have long been a holy grail in biology.
Steady progress was made for a long time (as judged by the CASP contest that has been taking place every 2 years since 1994), but the biggest breakthrough came when Deepmind’s AlphaFold burst on the scene with much better results than anyone else, and then subsequently improved enough that the problem can be considered solved to a significant extent (though not fully).
Meta’s AI team has also been working on this problem, and the results are pretty impressive:
Researchers at Meta (formerly Facebook, headquartered in Menlo Park, California) have used AI to predict the structures of some 600 million proteins from bacteria, viruses and other microorganisms that haven’t been characterized.
“These are the structures we know the least about. These are incredibly mysterious proteins. I think they offer the potential for great insight into biology,” says Alexander Rives, the research lead of Meta AI’s protein team.
Interestingly, they did this using a large language model (LLM), but instead of training it on human languages like English and French, they trained it on the sequences of amino acids of known proteins.
It’s a good reminder that, as Claude Shannon would say, whatever’s in the physical world can be represented as information.
Our software can be trained to understand this information that is isomorphic to the physical world, and so indirectly learns to understand the physical world.
When you think about it, this isn’t so different from how brains, which are stuck in our skulls and only get information about the outside indirectly through nerve signals.
One benefit of Meta’s approach seems to be speed (at the expense of precision, sometimes):
In total, the team predicted the structures of more than 617 million proteins. The effort took just two weeks (by contrast, AlphaFold can take minutes to generate a single prediction). The structures are freely available for use, as is the code underlying the model, says Rives.
Of the 617 million predictions, the model deemed more than one-third to be high quality, such that researchers can have confidence that the overall protein shape is correct and, in some cases, can discern atomic-level details. Millions of these structures are entirely unlike anything in the databases of protein structures determined experimentally, or any of AlphaFold’s predictions from known organisms.
💾 💾 🛩 Airlines still big purchasers of floppy disks 💾 💾 🏭
There are few manufacturers of floppy disks left. If you ask one of the last ones who buys them:
According to Persky, his main customers are the industrial users whom they define as the people using floppy disks as an agent for getting information in and out of a machine.
He says that the air fleets still use floppy disks as half of them are older than twenty years; and back then, disks were the best technology available.
“That’s a huge consumer. There’s also medical equipment, which requires floppy disks to get the information in and out of medical devices,” he adds.
Some things never die.
There are plenty of mainframe computers still in use that probably outlived the original technicians that maintained them. Thankfully fax machines are fading fast, but that took much longer than I would’ve expected…
How long can floppies stick around? Are nuclear codes in B-52 bombers on floppies?
h/t friend-of-the-show Doomberg (🟩🐓)
The Arts & History
📺 TV Show Episode Ratings Visualization
What to Watch on TV is a fun & useful site where you can see per-episode IMDB ratings for TV shows. It makes it very clear when certain shows jump the shark 🦈 or have widely disliked finales…